Init commit for swbd

egs/swbd/ASR/README.md

@@ -0,0 +1,29 @@
+# Switchboard
+
+The Switchboard-1 Telephone Speech Corpus (LDC97S62) consists of approximately 260 hours of speech and was originally collected by Texas Instruments in 1990-1, under DARPA sponsorship. The first release of the corpus was published by NIST and distributed by the LDC in 1992-3. Since that release, a number of corrections have been made to the data files as presented on the original CD-ROM set and all copies of the first pressing have been distributed.
+
+Switchboard is a collection of about 2,400 two-sided telephone conversations among 543 speakers (302 male, 241 female) from all areas of the United States. A computer-driven robot operator system handled the calls, giving the caller appropriate recorded prompts, selecting and dialing another person (the callee) to take part in a conversation, introducing a topic for discussion and recording the speech from the two subjects into separate channels until the conversation was finished. About 70 topics were provided, of which about 50 were used frequently. Selection of topics and callees was constrained so that: (1) no two speakers would converse together more than once and (2) no one spoke more than once on a given topic.
+
+(The above introduction is from the [LDC Switchboard-1 Release 2 webpage](https://catalog.ldc.upenn.edu/LDC97S62).)
+
+**Caution**: The `conformer_ctc` recipe for Switchboard is currently very rough and has a high Word Error Rate, requiring more improvement and refinement. The TODO list for this recipe is as follows.
+
+## TODO List
+- [ ] Incorporate Lhotse for dataprocessing
+- [ ] Refer to Global Mapping Rules when computing Word Error Rate
+- [ ] Detailed Word Error Rate summary for eval2000 (callhome, swbd) and rt03 (fsh, swbd) testset
+- [ ] Switchboard transcript train/dev split for LM training
+- [ ] Fisher corpus LDC2004T19 LDC2005T19 LDC2004S13 LDC2005S13 for LM training
+
+## Performance Record
+|                                |  eval2000  |  rt03  |
+|--------------------------------|------------|--------|
+|         `conformer_ctc`        |    33.37   |  35.06 |
+
+See [RESULTS](/egs/swbd/ASR/RESULTS.md) for details.
+
+## Credit
+
+The training script for `conformer_ctc` comes from the LibriSpeech `conformer_ctc` recipe in icefall.
+
+A lot of the scripts for data processing are from the first-gen Kaldi and the ESPNet project, tailored to incorporate with Lhotse and icefall.

egs/swbd/ASR/RESULTS.md

@@ -0,0 +1,50 @@
+## Results
+### Switchboard BPE training results (Conformer-CTC)
+
+#### 2023-06-26
+
+The best WER, as of 2023-06-26, for the Switchboard is below
+
+Results using HLG decoding + n-gram LM rescoring + attention decoder rescoring:
+
+|                                |  eval2000  |  rt03  |
+|--------------------------------|------------|--------|
+|         `conformer_ctc`        |    33.37   |  35.06 |
+
+Scale values used in n-gram LM rescoring and attention rescoring for the best WERs are:
+
+##### eval2000
+
+| ngram_lm_scale | attention_scale |
+|----------------|-----------------|
+|      0.3       |       2.5       |
+
+##### rt03
+
+| ngram_lm_scale | attention_scale |
+|----------------|-----------------|
+|      0.7       |       1.3       |
+
+To reproduce the above result, use the following commands for training:
+
+```bash
+cd egs/swbd/ASR
+./prepare.sh
+export CUDA_VISIBLE_DEVICES="0,1"
+./conformer_ctc/train.py \
+  --max-duration 120 \
+  --num-workers 8 \
+  --enable-musan False \
+  --world-size 2 \
+```
+
+and the following command for decoding:
+
+```bash
+./conformer_ctc/decode.py \
+  --epoch 55 \
+  --avg 1 \
+  --max-duration 50
+```
+
+

egs/swbd/ASR/conformer_ctc/ali.py

@@ -0,0 +1,395 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Usage:
+    ./conformer_ctc/ali.py \
+            --exp-dir ./conformer_ctc/exp \
+            --lang-dir ./data/lang_bpe_500 \
+            --epoch 20 \
+            --avg 10 \
+            --max-duration 300 \
+            --dataset train-clean-100 \
+            --out-dir data/ali
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import k2
+import numpy as np
+import torch
+from asr_datamodule import LibriSpeechAsrDataModule
+from conformer import Conformer
+from lhotse import CutSet
+from lhotse.features.io import FeaturesWriter, NumpyHdf5Writer
+
+from icefall.bpe_graph_compiler import BpeCtcTrainingGraphCompiler
+from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.decode import one_best_decoding
+from icefall.env import get_env_info
+from icefall.lexicon import Lexicon
+from icefall.utils import (
+    AttributeDict,
+    encode_supervisions,
+    get_alignments,
+    setup_logger,
+)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=34,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=20,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_bpe_500",
+        help="The lang dir",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="conformer_ctc/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--out-dir",
+        type=str,
+        required=True,
+        help="""Output directory.
+        It contains 3 generated files:
+
+        - labels_xxx.h5
+        - aux_labels_xxx.h5
+        - librispeech_cuts_xxx.jsonl.gz
+
+        where xxx is the value of `--dataset`. For instance, if
+        `--dataset` is `train-clean-100`, it will contain 3 files:
+
+        - `labels_train-clean-100.h5`
+        - `aux_labels_train-clean-100.h5`
+        - `librispeech_cuts_train-clean-100.jsonl.gz`
+
+        Note: Both labels_xxx.h5 and aux_labels_xxx.h5 contain framewise
+        alignment. The difference is that labels_xxx.h5 contains repeats.
+        """,
+    )
+
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        required=True,
+        help="""The name of the dataset to compute alignments for.
+        Possible values are:
+            - test-clean.
+            - test-other
+            - train-clean-100
+            - train-clean-360
+            - train-other-500
+            - dev-clean
+            - dev-other
+        """,
+    )
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            "lm_dir": Path("data/lm"),
+            "feature_dim": 80,
+            "nhead": 8,
+            "attention_dim": 512,
+            "subsampling_factor": 4,
+            # Set it to 0 since attention decoder
+            # is not used for computing alignments
+            "num_decoder_layers": 0,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            "output_beam": 10,
+            "use_double_scores": True,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def compute_alignments(
+    model: torch.nn.Module,
+    dl: torch.utils.data.DataLoader,
+    labels_writer: FeaturesWriter,
+    aux_labels_writer: FeaturesWriter,
+    params: AttributeDict,
+    graph_compiler: BpeCtcTrainingGraphCompiler,
+) -> CutSet:
+    """Compute the framewise alignments of a dataset.
+
+    Args:
+      model:
+        The neural network model.
+      dl:
+        Dataloader containing the dataset.
+      params:
+        Parameters for computing alignments.
+      graph_compiler:
+        It converts token IDs to decoding graphs.
+    Returns:
+      Return a CutSet. Each cut has two custom fields: labels_alignment
+      and aux_labels_alignment, containing framewise alignments information.
+      Both are of type `lhotse.array.TemporalArray`. The difference between
+      the two alignments is that `labels_alignment` contain repeats.
+    """
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+    num_cuts = 0
+
+    device = graph_compiler.device
+    cuts = []
+    for batch_idx, batch in enumerate(dl):
+        feature = batch["inputs"]
+
+        # at entry, feature is [N, T, C]
+        assert feature.ndim == 3
+        feature = feature.to(device)
+
+        supervisions = batch["supervisions"]
+        cut_list = supervisions["cut"]
+
+        for cut in cut_list:
+            assert len(cut.supervisions) == 1, f"{len(cut.supervisions)}"
+
+        nnet_output, encoder_memory, memory_mask = model(feature, supervisions)
+        # nnet_output is [N, T, C]
+        supervision_segments, texts = encode_supervisions(
+            supervisions, subsampling_factor=params.subsampling_factor
+        )
+        # we need also to sort cut_ids as encode_supervisions()
+        # reorders "texts".
+        # In general, new2old is an identity map since lhotse sorts the returned
+        # cuts by duration in descending order
+        new2old = supervision_segments[:, 0].tolist()
+
+        cut_list = [cut_list[i] for i in new2old]
+
+        token_ids = graph_compiler.texts_to_ids(texts)
+        decoding_graph = graph_compiler.compile(token_ids)
+
+        dense_fsa_vec = k2.DenseFsaVec(
+            nnet_output,
+            supervision_segments,
+            allow_truncate=params.subsampling_factor - 1,
+        )
+
+        lattice = k2.intersect_dense(
+            decoding_graph,
+            dense_fsa_vec,
+            params.output_beam,
+        )
+
+        best_path = one_best_decoding(
+            lattice=lattice,
+            use_double_scores=params.use_double_scores,
+        )
+
+        labels_ali = get_alignments(best_path, kind="labels")
+        aux_labels_ali = get_alignments(best_path, kind="aux_labels")
+        assert len(labels_ali) == len(aux_labels_ali) == len(cut_list)
+        for cut, labels, aux_labels in zip(cut_list, labels_ali, aux_labels_ali):
+            cut.labels_alignment = labels_writer.store_array(
+                key=cut.id,
+                value=np.asarray(labels, dtype=np.int32),
+                # frame shift is 0.01s, subsampling_factor is 4
+                frame_shift=0.04,
+                temporal_dim=0,
+                start=0,
+            )
+            cut.aux_labels_alignment = aux_labels_writer.store_array(
+                key=cut.id,
+                value=np.asarray(aux_labels, dtype=np.int32),
+                # frame shift is 0.01s, subsampling_factor is 4
+                frame_shift=0.04,
+                temporal_dim=0,
+                start=0,
+            )
+
+        cuts += cut_list
+
+        num_cuts += len(cut_list)
+
+        if batch_idx % 100 == 0:
+            batch_str = f"{batch_idx}/{num_batches}"
+
+            logging.info(f"batch {batch_str}, cuts processed until now is {num_cuts}")
+
+    return CutSet.from_cuts(cuts)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+
+    args.enable_spec_aug = False
+    args.enable_musan = False
+    args.return_cuts = True
+    args.concatenate_cuts = False
+
+    params = get_params()
+    params.update(vars(args))
+
+    setup_logger(f"{params.exp_dir}/log-ali")
+
+    logging.info(f"Computing alignments for {params.dataset} - started")
+    logging.info(params)
+
+    out_dir = Path(params.out_dir)
+    out_dir.mkdir(exist_ok=True)
+
+    out_labels_ali_filename = out_dir / f"labels_{params.dataset}.h5"
+    out_aux_labels_ali_filename = out_dir / f"aux_labels_{params.dataset}.h5"
+    out_manifest_filename = out_dir / f"librispeech_cuts_{params.dataset}.jsonl.gz"
+
+    for f in (
+        out_labels_ali_filename,
+        out_aux_labels_ali_filename,
+        out_manifest_filename,
+    ):
+        if f.exists():
+            logging.info(f"{f} exists - skipping")
+            return
+
+    lexicon = Lexicon(params.lang_dir)
+    max_token_id = max(lexicon.tokens)
+    num_classes = max_token_id + 1  # +1 for the blank
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+    logging.info(f"device: {device}")
+
+    graph_compiler = BpeCtcTrainingGraphCompiler(
+        params.lang_dir,
+        device=device,
+        sos_token="<sos/eos>",
+        eos_token="<sos/eos>",
+    )
+
+    logging.info("About to create model")
+    model = Conformer(
+        num_features=params.feature_dim,
+        nhead=params.nhead,
+        d_model=params.attention_dim,
+        num_classes=num_classes,
+        subsampling_factor=params.subsampling_factor,
+        num_decoder_layers=params.num_decoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    model.to(device)
+
+    if params.avg == 1:
+        load_checkpoint(
+            f"{params.exp_dir}/epoch-{params.epoch}.pt", model, strict=False
+        )
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.load_state_dict(
+            average_checkpoints(filenames, device=device), strict=False
+        )
+
+    model.eval()
+
+    librispeech = LibriSpeechAsrDataModule(args)
+    if params.dataset == "test-clean":
+        test_clean_cuts = librispeech.test_clean_cuts()
+        dl = librispeech.test_dataloaders(test_clean_cuts)
+    elif params.dataset == "test-other":
+        test_other_cuts = librispeech.test_other_cuts()
+        dl = librispeech.test_dataloaders(test_other_cuts)
+    elif params.dataset == "train-clean-100":
+        train_clean_100_cuts = librispeech.train_clean_100_cuts()
+        dl = librispeech.train_dataloaders(train_clean_100_cuts)
+    elif params.dataset == "train-clean-360":
+        train_clean_360_cuts = librispeech.train_clean_360_cuts()
+        dl = librispeech.train_dataloaders(train_clean_360_cuts)
+    elif params.dataset == "train-other-500":
+        train_other_500_cuts = librispeech.train_other_500_cuts()
+        dl = librispeech.train_dataloaders(train_other_500_cuts)
+    elif params.dataset == "dev-clean":
+        dev_clean_cuts = librispeech.dev_clean_cuts()
+        dl = librispeech.valid_dataloaders(dev_clean_cuts)
+    else:
+        assert params.dataset == "dev-other", f"{params.dataset}"
+        dev_other_cuts = librispeech.dev_other_cuts()
+        dl = librispeech.valid_dataloaders(dev_other_cuts)
+
+    logging.info(f"Processing {params.dataset}")
+    with NumpyHdf5Writer(out_labels_ali_filename) as labels_writer:
+        with NumpyHdf5Writer(out_aux_labels_ali_filename) as aux_labels_writer:
+            cut_set = compute_alignments(
+                model=model,
+                dl=dl,
+                labels_writer=labels_writer,
+                aux_labels_writer=aux_labels_writer,
+                params=params,
+                graph_compiler=graph_compiler,
+            )
+
+    cut_set.to_file(out_manifest_filename)
+
+    logging.info(
+        f"For dataset {params.dataset}, its alignments with repeats are "
+        f"saved to {out_labels_ali_filename}, the alignments without repeats "
+        f"are saved to {out_aux_labels_ali_filename}, and the cut manifest "
+        f"file is {out_manifest_filename}. Number of cuts: {len(cut_set)}"
+    )
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/conformer_ctc/asr_datamodule.py

@@ -0,0 +1,410 @@
+# Copyright      2021  Piotr Żelasko
+# Copyright      2022  Xiaomi Corporation     (Author: Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import argparse
+import inspect
+import logging
+from functools import lru_cache
+from pathlib import Path
+from typing import Any, Dict, Optional
+
+import torch
+from lhotse import CutSet, Fbank, FbankConfig, load_manifest, load_manifest_lazy
+from lhotse.dataset import (  # noqa F401 for PrecomputedFeatures
+    CutConcatenate,
+    CutMix,
+    DynamicBucketingSampler,
+    K2SpeechRecognitionDataset,
+    PrecomputedFeatures,
+    SingleCutSampler,
+    SpecAugment,
+)
+from lhotse.dataset.input_strategies import (  # noqa F401 For AudioSamples
+    AudioSamples,
+    OnTheFlyFeatures,
+)
+from lhotse.utils import fix_random_seed
+from torch.utils.data import DataLoader
+
+from icefall.utils import str2bool
+
+
+class _SeedWorkers:
+    def __init__(self, seed: int):
+        self.seed = seed
+
+    def __call__(self, worker_id: int):
+        fix_random_seed(self.seed + worker_id)
+
+
+class SwitchBoardAsrDataModule:
+    """
+    DataModule for k2 ASR experiments.
+    It assumes there is always one train dataloader,
+    but there can be multiple test dataloaders (e.g. SwitchBoard rt03
+    and eval2000).
+
+    It contains all the common data pipeline modules used in ASR
+    experiments, e.g.:
+    - dynamic batch size,
+    - bucketing samplers,
+    - cut concatenation,
+    - augmentation,
+    - on-the-fly feature extraction
+
+    This class should be derived for specific corpora used in ASR tasks.
+    """
+
+    def __init__(self, args: argparse.Namespace):
+        self.args = args
+
+    @classmethod
+    def add_arguments(cls, parser: argparse.ArgumentParser):
+        group = parser.add_argument_group(
+            title="ASR data related options",
+            description="These options are used for the preparation of "
+            "PyTorch DataLoaders from Lhotse CutSet's -- they control the "
+            "effective batch sizes, sampling strategies, applied data "
+            "augmentations, etc.",
+        )
+        group.add_argument(
+            "--manifest-dir",
+            type=Path,
+            default=Path("data/fbank"),
+            help="Path to directory with train/valid/test cuts.",
+        )
+        group.add_argument(
+            "--max-duration",
+            type=int,
+            default=200.0,
+            help="Maximum pooled recordings duration (seconds) in a "
+            "single batch. You can reduce it if it causes CUDA OOM.",
+        )
+        group.add_argument(
+            "--bucketing-sampler",
+            type=str2bool,
+            default=True,
+            help="When enabled, the batches will come from buckets of "
+            "similar duration (saves padding frames).",
+        )
+        group.add_argument(
+            "--num-buckets",
+            type=int,
+            default=30,
+            help="The number of buckets for the DynamicBucketingSampler"
+            "(you might want to increase it for larger datasets).",
+        )
+        group.add_argument(
+            "--concatenate-cuts",
+            type=str2bool,
+            default=False,
+            help="When enabled, utterances (cuts) will be concatenated "
+            "to minimize the amount of padding.",
+        )
+        group.add_argument(
+            "--duration-factor",
+            type=float,
+            default=1.0,
+            help="Determines the maximum duration of a concatenated cut "
+            "relative to the duration of the longest cut in a batch.",
+        )
+        group.add_argument(
+            "--gap",
+            type=float,
+            default=1.0,
+            help="The amount of padding (in seconds) inserted between "
+            "concatenated cuts. This padding is filled with noise when "
+            "noise augmentation is used.",
+        )
+        group.add_argument(
+            "--on-the-fly-feats",
+            type=str2bool,
+            default=False,
+            help="When enabled, use on-the-fly cut mixing and feature "
+            "extraction. Will drop existing precomputed feature manifests "
+            "if available.",
+        )
+        group.add_argument(
+            "--shuffle",
+            type=str2bool,
+            default=True,
+            help="When enabled (=default), the examples will be "
+            "shuffled for each epoch.",
+        )
+        group.add_argument(
+            "--drop-last",
+            type=str2bool,
+            default=True,
+            help="Whether to drop last batch. Used by sampler.",
+        )
+        group.add_argument(
+            "--return-cuts",
+            type=str2bool,
+            default=True,
+            help="When enabled, each batch will have the "
+            "field: batch['supervisions']['cut'] with the cuts that "
+            "were used to construct it.",
+        )
+
+        group.add_argument(
+            "--num-workers",
+            type=int,
+            default=2,
+            help="The number of training dataloader workers that "
+            "collect the batches.",
+        )
+
+        group.add_argument(
+            "--enable-spec-aug",
+            type=str2bool,
+            default=True,
+            help="When enabled, use SpecAugment for training dataset.",
+        )
+
+        group.add_argument(
+            "--spec-aug-time-warp-factor",
+            type=int,
+            default=80,
+            help="Used only when --enable-spec-aug is True. "
+            "It specifies the factor for time warping in SpecAugment. "
+            "Larger values mean more warping. "
+            "A value less than 1 means to disable time warp.",
+        )
+
+        group.add_argument(
+            "--enable-musan",
+            type=str2bool,
+            default=True,
+            help="When enabled, select noise from MUSAN and mix it"
+            "with training dataset. ",
+        )
+
+        group.add_argument(
+            "--input-strategy",
+            type=str,
+            default="PrecomputedFeatures",
+            help="AudioSamples or PrecomputedFeatures",
+        )
+
+    def train_dataloaders(
+        self,
+        cuts_train: CutSet,
+        sampler_state_dict: Optional[Dict[str, Any]] = None,
+    ) -> DataLoader:
+        """
+        Args:
+          cuts_train:
+            CutSet for training.
+          sampler_state_dict:
+            The state dict for the training sampler.
+        """
+        transforms = []
+        if self.args.enable_musan:
+            logging.info("Enable MUSAN")
+            logging.info("About to get Musan cuts")
+            cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz")
+            transforms.append(
+                CutMix(cuts=cuts_musan, prob=0.5, snr=(10, 20), preserve_id=True)
+            )
+        else:
+            logging.info("Disable MUSAN")
+
+        if self.args.concatenate_cuts:
+            logging.info(
+                f"Using cut concatenation with duration factor "
+                f"{self.args.duration_factor} and gap {self.args.gap}."
+            )
+            # Cut concatenation should be the first transform in the list,
+            # so that if we e.g. mix noise in, it will fill the gaps between
+            # different utterances.
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        input_transforms = []
+        if self.args.enable_spec_aug:
+            logging.info("Enable SpecAugment")
+            logging.info(f"Time warp factor: {self.args.spec_aug_time_warp_factor}")
+            # Set the value of num_frame_masks according to Lhotse's version.
+            # In different Lhotse's versions, the default of num_frame_masks is
+            # different.
+            num_frame_masks = 10
+            num_frame_masks_parameter = inspect.signature(
+                SpecAugment.__init__
+            ).parameters["num_frame_masks"]
+            if num_frame_masks_parameter.default == 1:
+                num_frame_masks = 2
+            logging.info(f"Num frame mask: {num_frame_masks}")
+            input_transforms.append(
+                SpecAugment(
+                    time_warp_factor=self.args.spec_aug_time_warp_factor,
+                    num_frame_masks=num_frame_masks,
+                    features_mask_size=27,
+                    num_feature_masks=2,
+                    frames_mask_size=100,
+                )
+            )
+        else:
+            logging.info("Disable SpecAugment")
+
+        logging.info("About to create train dataset")
+        train = K2SpeechRecognitionDataset(
+            input_strategy=eval(self.args.input_strategy)(),
+            cut_transforms=transforms,
+            input_transforms=input_transforms,
+            return_cuts=self.args.return_cuts,
+        )
+
+        if self.args.on_the_fly_feats:
+            # NOTE: the PerturbSpeed transform should be added only if we
+            # remove it from data prep stage.
+            # Add on-the-fly speed perturbation; since originally it would
+            # have increased epoch size by 3, we will apply prob 2/3 and use
+            # 3x more epochs.
+            # Speed perturbation probably should come first before
+            # concatenation, but in principle the transforms order doesn't have
+            # to be strict (e.g. could be randomized)
+            # transforms = [PerturbSpeed(factors=[0.9, 1.1], p=2/3)] + transforms   # noqa
+            # Drop feats to be on the safe side.
+            train = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80))),
+                input_transforms=input_transforms,
+                return_cuts=self.args.return_cuts,
+            )
+
+        if self.args.bucketing_sampler:
+            logging.info("Using DynamicBucketingSampler.")
+            train_sampler = DynamicBucketingSampler(
+                cuts_train,
+                max_duration=self.args.max_duration,
+                shuffle=self.args.shuffle,
+                num_buckets=self.args.num_buckets,
+                drop_last=self.args.drop_last,
+            )
+        else:
+            logging.info("Using SingleCutSampler.")
+            train_sampler = SingleCutSampler(
+                cuts_train,
+                max_duration=self.args.max_duration,
+                shuffle=self.args.shuffle,
+            )
+        logging.info("About to create train dataloader")
+
+        if sampler_state_dict is not None:
+            logging.info("Loading sampler state dict")
+            train_sampler.load_state_dict(sampler_state_dict)
+
+        # 'seed' is derived from the current random state, which will have
+        # previously been set in the main process.
+        seed = torch.randint(0, 100000, ()).item()
+        worker_init_fn = _SeedWorkers(seed)
+
+        train_dl = DataLoader(
+            train,
+            sampler=train_sampler,
+            batch_size=None,
+            num_workers=self.args.num_workers,
+            persistent_workers=False,
+            worker_init_fn=worker_init_fn,
+        )
+
+        return train_dl
+
+    def valid_dataloaders(self, cuts_valid: CutSet) -> DataLoader:
+        transforms = []
+        if self.args.concatenate_cuts:
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        logging.info("About to create dev dataset")
+        if self.args.on_the_fly_feats:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80))),
+                return_cuts=self.args.return_cuts,
+            )
+        else:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                return_cuts=self.args.return_cuts,
+            )
+        valid_sampler = DynamicBucketingSampler(
+            cuts_valid,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+        )
+        logging.info("About to create dev dataloader")
+        valid_dl = DataLoader(
+            validate,
+            sampler=valid_sampler,
+            batch_size=None,
+            num_workers=2,
+            persistent_workers=False,
+        )
+
+        return valid_dl
+
+    def test_dataloaders(self, cuts: CutSet) -> DataLoader:
+        logging.debug("About to create test dataset")
+        test = K2SpeechRecognitionDataset(
+            input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80)))
+            if self.args.on_the_fly_feats
+            else eval(self.args.input_strategy)(),
+            return_cuts=self.args.return_cuts,
+        )
+        sampler = DynamicBucketingSampler(
+            cuts,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+        )
+        logging.debug("About to create test dataloader")
+        test_dl = DataLoader(
+            test,
+            batch_size=None,
+            sampler=sampler,
+            num_workers=self.args.num_workers,
+        )
+        return test_dl
+
+    @lru_cache()
+    def train_all_cuts(self) -> CutSet:
+        logging.info("switchboard: About to get train cuts")
+        return load_manifest_lazy(self.args.manifest_dir / "swbd_cuts_all.jsonl.gz").subset(last=2388).trim_to_supervisions(keep_all_channels=True)
+
+    @lru_cache()
+    def dev_cuts(self) -> CutSet:
+        logging.info("switchboard: About to get dev cuts")
+        return load_manifest_lazy(self.args.manifest_dir / "swbd_cuts_all.jsonl.gz").subset(first=50).trim_to_supervisions(keep_all_channels=True)
+
+    @lru_cache()
+    def test_eval2000_cuts(self) -> CutSet:
+        logging.info("switchboard: About to get eval2000 cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "swbd_cuts_eval2000.jsonl.gz"
+        )
+
+    @lru_cache()
+    def test_rt03_cuts(self) -> CutSet:
+        logging.info("switchboard: About to get rt03 cuts")
+        return load_manifest_lazy(self.args.manifest_dir / "swbd_cuts_rt03.jsonl.gz")

egs/swbd/ASR/conformer_ctc/conformer.py

@@ -0,0 +1,910 @@
+#!/usr/bin/env python3
+# Copyright (c)  2021  University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import math
+import warnings
+from typing import Optional, Tuple, Union
+
+import torch
+from torch import Tensor, nn
+from transformer import Supervisions, Transformer, encoder_padding_mask
+
+
+class Conformer(Transformer):
+    """
+    Args:
+        num_features (int): Number of input features
+        num_classes (int): Number of output classes
+        subsampling_factor (int): subsampling factor of encoder (the convolution layers before transformers)
+        d_model (int): attention dimension
+        nhead (int): number of head
+        dim_feedforward (int): feedforward dimention
+        num_encoder_layers (int): number of encoder layers
+        num_decoder_layers (int): number of decoder layers
+        dropout (float): dropout rate
+        cnn_module_kernel (int): Kernel size of convolution module
+        normalize_before (bool): whether to use layer_norm before the first block.
+        vgg_frontend (bool): whether to use vgg frontend.
+    """
+
+    def __init__(
+        self,
+        num_features: int,
+        num_classes: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        num_decoder_layers: int = 6,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: Union[float, bool] = 0.1,
+    ) -> None:
+        super(Conformer, self).__init__(
+            num_features=num_features,
+            num_classes=num_classes,
+            subsampling_factor=subsampling_factor,
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            num_encoder_layers=num_encoder_layers,
+            num_decoder_layers=num_decoder_layers,
+            dropout=dropout,
+            normalize_before=normalize_before,
+            vgg_frontend=vgg_frontend,
+            use_feat_batchnorm=use_feat_batchnorm,
+        )
+
+        self.encoder_pos = RelPositionalEncoding(d_model, dropout)
+
+        use_conv_batchnorm = True
+        if isinstance(use_feat_batchnorm, float):
+            use_conv_batchnorm = False
+        encoder_layer = ConformerEncoderLayer(
+            d_model,
+            nhead,
+            dim_feedforward,
+            dropout,
+            cnn_module_kernel,
+            normalize_before,
+            use_conv_batchnorm,
+        )
+        self.encoder = ConformerEncoder(encoder_layer, num_encoder_layers)
+        self.normalize_before = normalize_before
+        if self.normalize_before:
+            self.after_norm = nn.LayerNorm(d_model)
+        else:
+            # Note: TorchScript detects that self.after_norm could be used inside forward()
+            #       and throws an error without this change.
+            self.after_norm = identity
+
+    def run_encoder(
+        self, x: Tensor, supervisions: Optional[Supervisions] = None
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        """
+        Args:
+          x:
+            The model input. Its shape is (N, T, C).
+          supervisions:
+            Supervision in lhotse format.
+            See https://github.com/lhotse-speech/lhotse/blob/master/lhotse/dataset/speech_recognition.py#L32  # noqa
+            CAUTION: It contains length information, i.e., start and number of
+            frames, before subsampling
+            It is read directly from the batch, without any sorting. It is used
+            to compute encoder padding mask, which is used as memory key padding
+            mask for the decoder.
+
+        Returns:
+            Tensor: Predictor tensor of dimension (input_length, batch_size, d_model).
+            Tensor: Mask tensor of dimension (batch_size, input_length)
+        """
+        x = self.encoder_embed(x)
+        x, pos_emb = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (B, T, F) -> (T, B, F)
+        mask = encoder_padding_mask(x.size(0), supervisions)
+        if mask is not None:
+            mask = mask.to(x.device)
+        x = self.encoder(x, pos_emb, src_key_padding_mask=mask)  # (T, B, F)
+
+        if self.normalize_before:
+            x = self.after_norm(x)
+
+        return x, mask
+
+
+class ConformerEncoderLayer(nn.Module):
+    """
+    ConformerEncoderLayer is made up of self-attn, feedforward and convolution networks.
+    See: "Conformer: Convolution-augmented Transformer for Speech Recognition"
+
+    Args:
+        d_model: the number of expected features in the input (required).
+        nhead: the number of heads in the multiheadattention models (required).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+        cnn_module_kernel (int): Kernel size of convolution module.
+        normalize_before: whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = encoder_layer(src, pos_emb)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+        use_conv_batchnorm: bool = False,
+    ) -> None:
+        super(ConformerEncoderLayer, self).__init__()
+        self.self_attn = RelPositionMultiheadAttention(d_model, nhead, dropout=0.0)
+
+        self.feed_forward = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.feed_forward_macaron = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.conv_module = ConvolutionModule(
+            d_model, cnn_module_kernel, use_batchnorm=use_conv_batchnorm
+        )
+
+        self.norm_ff_macaron = nn.LayerNorm(d_model)  # for the macaron style FNN module
+        self.norm_ff = nn.LayerNorm(d_model)  # for the FNN module
+        self.norm_mha = nn.LayerNorm(d_model)  # for the MHA module
+
+        self.ff_scale = 0.5
+
+        self.norm_conv = nn.LayerNorm(d_model)  # for the CNN module
+        self.norm_final = nn.LayerNorm(d_model)  # for the final output of the block
+
+        self.dropout = nn.Dropout(dropout)
+
+        self.normalize_before = normalize_before
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        src_mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            pos_emb: Positional embedding tensor (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, N is the batch size, E is the feature number
+        """
+
+        # macaron style feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff_macaron(src)
+        src = residual + self.ff_scale * self.dropout(self.feed_forward_macaron(src))
+        if not self.normalize_before:
+            src = self.norm_ff_macaron(src)
+
+        # multi-headed self-attention module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_mha(src)
+        src_att = self.self_attn(
+            src,
+            src,
+            src,
+            pos_emb=pos_emb,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout(src_att)
+        if not self.normalize_before:
+            src = self.norm_mha(src)
+
+        # convolution module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_conv(src)
+        src = residual + self.dropout(
+            self.conv_module(src, src_key_padding_mask=src_key_padding_mask)
+        )
+        if not self.normalize_before:
+            src = self.norm_conv(src)
+
+        # feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff(src)
+        src = residual + self.ff_scale * self.dropout(self.feed_forward(src))
+        if not self.normalize_before:
+            src = self.norm_ff(src)
+
+        if self.normalize_before:
+            src = self.norm_final(src)
+
+        return src
+
+
+class ConformerEncoder(nn.TransformerEncoder):
+    r"""ConformerEncoder is a stack of N encoder layers
+
+    Args:
+        encoder_layer: an instance of the ConformerEncoderLayer() class (required).
+        num_layers: the number of sub-encoder-layers in the encoder (required).
+        norm: the layer normalization component (optional).
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> conformer_encoder = ConformerEncoder(encoder_layer, num_layers=6)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = conformer_encoder(src, pos_emb)
+    """
+
+    def __init__(
+        self, encoder_layer: nn.Module, num_layers: int, norm: nn.Module = None
+    ) -> None:
+        super(ConformerEncoder, self).__init__(
+            encoder_layer=encoder_layer, num_layers=num_layers, norm=norm
+        )
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        r"""Pass the input through the encoder layers in turn.
+
+        Args:
+            src: the sequence to the encoder (required).
+            pos_emb: Positional embedding tensor (required).
+            mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length, N is the batch size, E is the feature number
+
+        """
+        output = src
+
+        for mod in self.layers:
+            output = mod(
+                output,
+                pos_emb,
+                src_mask=mask,
+                src_key_padding_mask=src_key_padding_mask,
+            )
+
+        if self.norm is not None:
+            output = self.norm(output)
+
+        return output
+
+
+class RelPositionalEncoding(torch.nn.Module):
+    """Relative positional encoding module.
+
+    See : Appendix B in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/embedding.py
+
+    Args:
+        d_model: Embedding dimension.
+        dropout_rate: Dropout rate.
+        max_len: Maximum input length.
+
+    """
+
+    def __init__(self, d_model: int, dropout_rate: float, max_len: int = 5000) -> None:
+        """Construct an PositionalEncoding object."""
+        super(RelPositionalEncoding, self).__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = torch.nn.Dropout(p=dropout_rate)
+        self.pe = None
+        self.extend_pe(torch.tensor(0.0).expand(1, max_len))
+
+    def extend_pe(self, x: Tensor) -> None:
+        """Reset the positional encodings."""
+        if self.pe is not None:
+            # self.pe contains both positive and negative parts
+            # the length of self.pe is 2 * input_len - 1
+            if self.pe.size(1) >= x.size(1) * 2 - 1:
+                # Note: TorchScript doesn't implement operator== for torch.Device
+                if self.pe.dtype != x.dtype or str(self.pe.device) != str(x.device):
+                    self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        # Suppose `i` means to the position of query vector and `j` means the
+        # position of key vector. We use position relative positions when keys
+        # are to the left (i>j) and negative relative positions otherwise (i<j).
+        pe_positive = torch.zeros(x.size(1), self.d_model)
+        pe_negative = torch.zeros(x.size(1), self.d_model)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe_positive[:, 0::2] = torch.sin(position * div_term)
+        pe_positive[:, 1::2] = torch.cos(position * div_term)
+        pe_negative[:, 0::2] = torch.sin(-1 * position * div_term)
+        pe_negative[:, 1::2] = torch.cos(-1 * position * div_term)
+
+        # Reserve the order of positive indices and concat both positive and
+        # negative indices. This is used to support the shifting trick
+        # as in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+        pe_positive = torch.flip(pe_positive, [0]).unsqueeze(0)
+        pe_negative = pe_negative[1:].unsqueeze(0)
+        pe = torch.cat([pe_positive, pe_negative], dim=1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> Tuple[Tensor, Tensor]:
+        """Add positional encoding.
+
+        Args:
+            x (torch.Tensor): Input tensor (batch, time, `*`).
+
+        Returns:
+            torch.Tensor: Encoded tensor (batch, time, `*`).
+            torch.Tensor: Encoded tensor (batch, 2*time-1, `*`).
+
+        """
+        self.extend_pe(x)
+        x = x * self.xscale
+        pos_emb = self.pe[
+            :,
+            self.pe.size(1) // 2
+            - x.size(1)
+            + 1 : self.pe.size(1) // 2  # noqa E203
+            + x.size(1),
+        ]
+        return self.dropout(x), self.dropout(pos_emb)
+
+
+class RelPositionMultiheadAttention(nn.Module):
+    r"""Multi-Head Attention layer with relative position encoding
+
+    See reference: "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+
+    Args:
+        embed_dim: total dimension of the model.
+        num_heads: parallel attention heads.
+        dropout: a Dropout layer on attn_output_weights. Default: 0.0.
+
+    Examples::
+
+        >>> rel_pos_multihead_attn = RelPositionMultiheadAttention(embed_dim, num_heads)
+        >>> attn_output, attn_output_weights = multihead_attn(query, key, value, pos_emb)
+    """
+
+    def __init__(
+        self,
+        embed_dim: int,
+        num_heads: int,
+        dropout: float = 0.0,
+    ) -> None:
+        super(RelPositionMultiheadAttention, self).__init__()
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.dropout = dropout
+        self.head_dim = embed_dim // num_heads
+        assert (
+            self.head_dim * num_heads == self.embed_dim
+        ), "embed_dim must be divisible by num_heads"
+
+        self.in_proj = nn.Linear(embed_dim, 3 * embed_dim, bias=True)
+        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=True)
+
+        # linear transformation for positional encoding.
+        self.linear_pos = nn.Linear(embed_dim, embed_dim, bias=False)
+        # these two learnable bias are used in matrix c and matrix d
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        self.pos_bias_u = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+        self.pos_bias_v = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+
+        self._reset_parameters()
+
+    def _reset_parameters(self) -> None:
+        nn.init.xavier_uniform_(self.in_proj.weight)
+        nn.init.constant_(self.in_proj.bias, 0.0)
+        nn.init.constant_(self.out_proj.bias, 0.0)
+
+        nn.init.xavier_uniform_(self.pos_bias_u)
+        nn.init.xavier_uniform_(self.pos_bias_v)
+
+    def forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. When given a binary mask and a value is True,
+                the corresponding value on the attention layer will be ignored. When given
+                a byte mask and a value is non-zero, the corresponding value on the attention
+                layer will be ignored
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            - Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the position
+            with the zero positions will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensure that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            is not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            - Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+        return self.multi_head_attention_forward(
+            query,
+            key,
+            value,
+            pos_emb,
+            self.embed_dim,
+            self.num_heads,
+            self.in_proj.weight,
+            self.in_proj.bias,
+            self.dropout,
+            self.out_proj.weight,
+            self.out_proj.bias,
+            training=self.training,
+            key_padding_mask=key_padding_mask,
+            need_weights=need_weights,
+            attn_mask=attn_mask,
+        )
+
+    def rel_shift(self, x: Tensor) -> Tensor:
+        """Compute relative positional encoding.
+
+        Args:
+            x: Input tensor (batch, head, time1, 2*time1-1).
+                time1 means the length of query vector.
+
+        Returns:
+            Tensor: tensor of shape (batch, head, time1, time2)
+          (note: time2 has the same value as time1, but it is for
+          the key, while time1 is for the query).
+        """
+        (batch_size, num_heads, time1, n) = x.shape
+        assert n == 2 * time1 - 1
+        # Note: TorchScript requires explicit arg for stride()
+        batch_stride = x.stride(0)
+        head_stride = x.stride(1)
+        time1_stride = x.stride(2)
+        n_stride = x.stride(3)
+        return x.as_strided(
+            (batch_size, num_heads, time1, time1),
+            (batch_stride, head_stride, time1_stride - n_stride, n_stride),
+            storage_offset=n_stride * (time1 - 1),
+        )
+
+    def multi_head_attention_forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        embed_dim_to_check: int,
+        num_heads: int,
+        in_proj_weight: Tensor,
+        in_proj_bias: Tensor,
+        dropout_p: float,
+        out_proj_weight: Tensor,
+        out_proj_bias: Tensor,
+        training: bool = True,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            embed_dim_to_check: total dimension of the model.
+            num_heads: parallel attention heads.
+            in_proj_weight, in_proj_bias: input projection weight and bias.
+            dropout_p: probability of an element to be zeroed.
+            out_proj_weight, out_proj_bias: the output projection weight and bias.
+            training: apply dropout if is ``True``.
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. This is an binary mask. When the value is True,
+                the corresponding value on the attention layer will be filled with -inf.
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` or :math:`(1, 2*L-1, E)` where L is the target sequence
+            length, N is the batch size, E is the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the zero positions
+            will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensures that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            are not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+
+        tgt_len, bsz, embed_dim = query.size()
+        assert embed_dim == embed_dim_to_check
+        assert key.size(0) == value.size(0) and key.size(1) == value.size(1)
+
+        head_dim = embed_dim // num_heads
+        assert (
+            head_dim * num_heads == embed_dim
+        ), "embed_dim must be divisible by num_heads"
+        scaling = float(head_dim) ** -0.5
+
+        if torch.equal(query, key) and torch.equal(key, value):
+            # self-attention
+            q, k, v = nn.functional.linear(query, in_proj_weight, in_proj_bias).chunk(
+                3, dim=-1
+            )
+
+        elif torch.equal(key, value):
+            # encoder-decoder attention
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            k, v = nn.functional.linear(key, _w, _b).chunk(2, dim=-1)
+
+        else:
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = embed_dim * 2
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            k = nn.functional.linear(key, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim * 2
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            v = nn.functional.linear(value, _w, _b)
+
+        if attn_mask is not None:
+            assert (
+                attn_mask.dtype == torch.float32
+                or attn_mask.dtype == torch.float64
+                or attn_mask.dtype == torch.float16
+                or attn_mask.dtype == torch.uint8
+                or attn_mask.dtype == torch.bool
+            ), "Only float, byte, and bool types are supported for attn_mask, not {}".format(
+                attn_mask.dtype
+            )
+            if attn_mask.dtype == torch.uint8:
+                warnings.warn(
+                    "Byte tensor for attn_mask is deprecated. Use bool tensor instead."
+                )
+                attn_mask = attn_mask.to(torch.bool)
+
+            if attn_mask.dim() == 2:
+                attn_mask = attn_mask.unsqueeze(0)
+                if list(attn_mask.size()) != [1, query.size(0), key.size(0)]:
+                    raise RuntimeError("The size of the 2D attn_mask is not correct.")
+            elif attn_mask.dim() == 3:
+                if list(attn_mask.size()) != [
+                    bsz * num_heads,
+                    query.size(0),
+                    key.size(0),
+                ]:
+                    raise RuntimeError("The size of the 3D attn_mask is not correct.")
+            else:
+                raise RuntimeError(
+                    "attn_mask's dimension {} is not supported".format(attn_mask.dim())
+                )
+            # attn_mask's dim is 3 now.
+
+        # convert ByteTensor key_padding_mask to bool
+        if key_padding_mask is not None and key_padding_mask.dtype == torch.uint8:
+            warnings.warn(
+                "Byte tensor for key_padding_mask is deprecated. Use bool tensor instead."
+            )
+            key_padding_mask = key_padding_mask.to(torch.bool)
+
+        q = q.contiguous().view(tgt_len, bsz, num_heads, head_dim)
+        k = k.contiguous().view(-1, bsz, num_heads, head_dim)
+        v = v.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)
+
+        src_len = k.size(0)
+
+        if key_padding_mask is not None:
+            assert key_padding_mask.size(0) == bsz, "{} == {}".format(
+                key_padding_mask.size(0), bsz
+            )
+            assert key_padding_mask.size(1) == src_len, "{} == {}".format(
+                key_padding_mask.size(1), src_len
+            )
+
+        q = q.transpose(0, 1)  # (batch, time1, head, d_k)
+
+        pos_emb_bsz = pos_emb.size(0)
+        assert pos_emb_bsz in (1, bsz)  # actually it is 1
+        p = self.linear_pos(pos_emb).view(pos_emb_bsz, -1, num_heads, head_dim)
+        p = p.transpose(1, 2)  # (batch, head, 2*time1-1, d_k)
+
+        q_with_bias_u = (q + self.pos_bias_u).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        q_with_bias_v = (q + self.pos_bias_v).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        # compute attention score
+        # first compute matrix a and matrix c
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        k = k.permute(1, 2, 3, 0)  # (batch, head, d_k, time2)
+        matrix_ac = torch.matmul(q_with_bias_u, k)  # (batch, head, time1, time2)
+
+        # compute matrix b and matrix d
+        matrix_bd = torch.matmul(
+            q_with_bias_v, p.transpose(-2, -1)
+        )  # (batch, head, time1, 2*time1-1)
+        matrix_bd = self.rel_shift(matrix_bd)
+
+        attn_output_weights = (
+            matrix_ac + matrix_bd
+        ) * scaling  # (batch, head, time1, time2)
+
+        attn_output_weights = attn_output_weights.view(bsz * num_heads, tgt_len, -1)
+
+        assert list(attn_output_weights.size()) == [
+            bsz * num_heads,
+            tgt_len,
+            src_len,
+        ]
+
+        if attn_mask is not None:
+            if attn_mask.dtype == torch.bool:
+                attn_output_weights.masked_fill_(attn_mask, float("-inf"))
+            else:
+                attn_output_weights += attn_mask
+
+        if key_padding_mask is not None:
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            attn_output_weights = attn_output_weights.masked_fill(
+                key_padding_mask.unsqueeze(1).unsqueeze(2),
+                float("-inf"),
+            )
+            attn_output_weights = attn_output_weights.view(
+                bsz * num_heads, tgt_len, src_len
+            )
+
+        attn_output_weights = nn.functional.softmax(attn_output_weights, dim=-1)
+        attn_output_weights = nn.functional.dropout(
+            attn_output_weights, p=dropout_p, training=training
+        )
+
+        attn_output = torch.bmm(attn_output_weights, v)
+        assert list(attn_output.size()) == [bsz * num_heads, tgt_len, head_dim]
+        attn_output = (
+            attn_output.transpose(0, 1).contiguous().view(tgt_len, bsz, embed_dim)
+        )
+        attn_output = nn.functional.linear(attn_output, out_proj_weight, out_proj_bias)
+
+        if need_weights:
+            # average attention weights over heads
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            return attn_output, attn_output_weights.sum(dim=1) / num_heads
+        else:
+            return attn_output, None
+
+
+class ConvolutionModule(nn.Module):
+    """ConvolutionModule in Conformer model.
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/conformer/convolution.py
+
+    Args:
+        channels (int): The number of channels of conv layers.
+        kernel_size (int): Kernerl size of conv layers.
+        bias (bool): Whether to use bias in conv layers (default=True).
+
+    """
+
+    def __init__(
+        self,
+        channels: int,
+        kernel_size: int,
+        bias: bool = True,
+        use_batchnorm: bool = False,
+    ) -> None:
+        """Construct an ConvolutionModule object."""
+        super(ConvolutionModule, self).__init__()
+        # kernerl_size should be a odd number for 'SAME' padding
+        assert (kernel_size - 1) % 2 == 0
+        self.use_batchnorm = use_batchnorm
+
+        self.pointwise_conv1 = nn.Conv1d(
+            channels,
+            2 * channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.depthwise_conv = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size,
+            stride=1,
+            padding=(kernel_size - 1) // 2,
+            groups=channels,
+            bias=bias,
+        )
+        if self.use_batchnorm:
+            self.norm = nn.BatchNorm1d(channels)
+        self.pointwise_conv2 = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.activation = Swish()
+
+    def forward(
+        self,
+        x: Tensor,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        """Compute convolution module.
+
+        Args:
+            x: Input tensor (#time, batch, channels).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Returns:
+            Tensor: Output tensor (#time, batch, channels).
+
+        """
+        # exchange the temporal dimension and the feature dimension
+        x = x.permute(1, 2, 0)  # (#batch, channels, time).
+
+        # GLU mechanism
+        x = self.pointwise_conv1(x)  # (batch, 2*channels, time)
+        x = nn.functional.glu(x, dim=1)  # (batch, channels, time)
+
+        # 1D Depthwise Conv
+        if src_key_padding_mask is not None:
+            x.masked_fill_(src_key_padding_mask.unsqueeze(1).expand_as(x), 0.0)
+        x = self.depthwise_conv(x)
+        if self.use_batchnorm:
+            x = self.norm(x)
+        x = self.activation(x)
+
+        x = self.pointwise_conv2(x)  # (batch, channel, time)
+
+        return x.permute(2, 0, 1)
+
+
+class Swish(torch.nn.Module):
+    """Construct an Swish object."""
+
+    def forward(self, x: Tensor) -> Tensor:
+        """Return Swich activation function."""
+        return x * torch.sigmoid(x)
+
+
+def identity(x):
+    return x

egs/swbd/ASR/conformer_ctc/decode.py

@@ -0,0 +1,815 @@
+#!/usr/bin/env python3
+# Copyright 2021 Xiaomi Corporation (Author: Liyong Guo, Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import argparse
+import logging
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import SwitchBoardAsrDataModule
+from conformer import Conformer
+
+from icefall.bpe_graph_compiler import BpeCtcTrainingGraphCompiler
+from icefall.checkpoint import load_checkpoint
+from icefall.decode import (
+    get_lattice,
+    nbest_decoding,
+    nbest_oracle,
+    one_best_decoding,
+    rescore_with_attention_decoder,
+    rescore_with_n_best_list,
+    rescore_with_rnn_lm,
+    rescore_with_whole_lattice,
+)
+from icefall.env import get_env_info
+from icefall.lexicon import Lexicon
+from icefall.rnn_lm.model import RnnLmModel
+from icefall.utils import (
+    AttributeDict,
+    get_texts,
+    load_averaged_model,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=77,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=55,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--method",
+        type=str,
+        default="attention-decoder",
+        help="""Decoding method.
+        Supported values are:
+            - (0) ctc-decoding. Use CTC decoding. It uses a sentence piece
+              model, i.e., lang_dir/bpe.model, to convert word pieces to words.
+              It needs neither a lexicon nor an n-gram LM.
+            - (1) 1best. Extract the best path from the decoding lattice as the
+              decoding result.
+            - (2) nbest. Extract n paths from the decoding lattice; the path
+              with the highest score is the decoding result.
+            - (3) nbest-rescoring. Extract n paths from the decoding lattice,
+              rescore them with an n-gram LM (e.g., a 4-gram LM), the path with
+              the highest score is the decoding result.
+            - (4) whole-lattice-rescoring. Rescore the decoding lattice with an
+              n-gram LM (e.g., a 4-gram LM), the best path of rescored lattice
+              is the decoding result.
+            - (5) attention-decoder. Extract n paths from the LM rescored
+              lattice, the path with the highest score is the decoding result.
+            - (6) rnn-lm. Rescoring with attention-decoder and RNN LM. We assume
+              you have trained an RNN LM using ./rnn_lm/train.py
+            - (7) nbest-oracle. Its WER is the lower bound of any n-best
+              rescoring method can achieve. Useful for debugging n-best
+              rescoring method.
+        """,
+    )
+
+    parser.add_argument(
+        "--num-paths",
+        type=int,
+        default=100,
+        help="""Number of paths for n-best based decoding method.
+        Used only when "method" is one of the following values:
+        nbest, nbest-rescoring, attention-decoder, rnn-lm, and nbest-oracle
+        """,
+    )
+
+    parser.add_argument(
+        "--nbest-scale",
+        type=float,
+        default=0.5,
+        help="""The scale to be applied to `lattice.scores`.
+        It's needed if you use any kinds of n-best based rescoring.
+        Used only when "method" is one of the following values:
+        nbest, nbest-rescoring, attention-decoder, rnn-lm, and nbest-oracle
+        A smaller value results in more unique paths.
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="conformer_ctc/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_bpe_500",
+        help="The lang dir",
+    )
+
+    parser.add_argument(
+        "--lm-dir",
+        type=str,
+        default="data/lm",
+        help="""The n-gram LM dir.
+        It should contain either G_4_gram.pt or G_4_gram.fst.txt
+        """,
+    )
+
+    parser.add_argument(
+        "--rnn-lm-exp-dir",
+        type=str,
+        default="rnn_lm/exp",
+        help="""Used only when --method is rnn-lm.
+        It specifies the path to RNN LM exp dir.
+        """,
+    )
+
+    parser.add_argument(
+        "--rnn-lm-epoch",
+        type=int,
+        default=7,
+        help="""Used only when --method is rnn-lm.
+        It specifies the checkpoint to use.
+        """,
+    )
+
+    parser.add_argument(
+        "--rnn-lm-avg",
+        type=int,
+        default=2,
+        help="""Used only when --method is rnn-lm.
+        It specifies the number of checkpoints to average.
+        """,
+    )
+
+    parser.add_argument(
+        "--rnn-lm-embedding-dim",
+        type=int,
+        default=2048,
+        help="Embedding dim of the model",
+    )
+
+    parser.add_argument(
+        "--rnn-lm-hidden-dim",
+        type=int,
+        default=2048,
+        help="Hidden dim of the model",
+    )
+
+    parser.add_argument(
+        "--rnn-lm-num-layers",
+        type=int,
+        default=4,
+        help="Number of RNN layers the model",
+    )
+    parser.add_argument(
+        "--rnn-lm-tie-weights",
+        type=str2bool,
+        default=False,
+        help="""True to share the weights between the input embedding layer and the
+        last output linear layer
+        """,
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            # parameters for conformer
+            "subsampling_factor": 4,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            "feature_dim": 80,
+            "nhead": 8,
+            "attention_dim": 512,
+            "num_decoder_layers": 6,
+            # parameters for decoding
+            "search_beam": 20,
+            "output_beam": 8,
+            "min_active_states": 30,
+            "max_active_states": 10000,
+            "use_double_scores": True,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    rnn_lm_model: Optional[nn.Module],
+    HLG: Optional[k2.Fsa],
+    H: Optional[k2.Fsa],
+    bpe_model: Optional[spm.SentencePieceProcessor],
+    batch: dict,
+    word_table: k2.SymbolTable,
+    sos_id: int,
+    eos_id: int,
+    G: Optional[k2.Fsa] = None,
+) -> Dict[str, List[List[str]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+
+        - key: It indicates the setting used for decoding. For example,
+               if no rescoring is used, the key is the string `no_rescore`.
+               If LM rescoring is used, the key is the string `lm_scale_xxx`,
+               where `xxx` is the value of `lm_scale`. An example key is
+               `lm_scale_0.7`
+        - value: It contains the decoding result. `len(value)` equals to
+                 batch size. `value[i]` is the decoding result for the i-th
+                 utterance in the given batch.
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+
+        - params.method is "1best", it uses 1best decoding without LM rescoring.
+        - params.method is "nbest", it uses nbest decoding without LM rescoring.
+        - params.method is "nbest-rescoring", it uses nbest LM rescoring.
+        - params.method is "whole-lattice-rescoring", it uses whole lattice LM
+          rescoring.
+
+      model:
+        The neural model.
+      rnn_lm_model:
+        The neural model for RNN LM.
+      HLG:
+        The decoding graph. Used only when params.method is NOT ctc-decoding.
+      H:
+        The ctc topo. Used only when params.method is ctc-decoding.
+      bpe_model:
+        The BPE model. Used only when params.method is ctc-decoding.
+      batch:
+        It is the return value from iterating
+        `lhotse.dataset.K2SpeechRecognitionDataset`. See its documentation
+        for the format of the `batch`.
+      word_table:
+        The word symbol table.
+      sos_id:
+        The token ID of the SOS.
+      eos_id:
+        The token ID of the EOS.
+      G:
+        An LM. It is not None when params.method is "nbest-rescoring"
+        or "whole-lattice-rescoring". In general, the G in HLG
+        is a 3-gram LM, while this G is a 4-gram LM.
+    Returns:
+      Return the decoding result. See above description for the format of
+      the returned dict. Note: If it decodes to nothing, then return None.
+    """
+    if HLG is not None:
+        device = HLG.device
+    else:
+        device = H.device
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+    feature = feature.to(device)
+    # at entry, feature is (N, T, C)
+
+    supervisions = batch["supervisions"]
+
+    nnet_output, memory, memory_key_padding_mask = model(feature, supervisions)
+    # nnet_output is (N, T, C)
+
+    supervision_segments = torch.stack(
+        (
+            supervisions["sequence_idx"],
+            supervisions["start_frame"] // params.subsampling_factor,
+            supervisions["num_frames"] // params.subsampling_factor,
+        ),
+        1,
+    ).to(torch.int32)
+
+    if H is None:
+        assert HLG is not None
+        decoding_graph = HLG
+    else:
+        assert HLG is None
+        assert bpe_model is not None
+        decoding_graph = H
+
+    lattice = get_lattice(
+        nnet_output=nnet_output,
+        decoding_graph=decoding_graph,
+        supervision_segments=supervision_segments,
+        search_beam=params.search_beam,
+        output_beam=params.output_beam,
+        min_active_states=params.min_active_states,
+        max_active_states=params.max_active_states,
+        subsampling_factor=params.subsampling_factor,
+    )
+
+    if params.method == "ctc-decoding":
+        best_path = one_best_decoding(
+            lattice=lattice, use_double_scores=params.use_double_scores
+        )
+        # Note: `best_path.aux_labels` contains token IDs, not word IDs
+        # since we are using H, not HLG here.
+        #
+        # token_ids is a lit-of-list of IDs
+        token_ids = get_texts(best_path)
+
+        # hyps is a list of str, e.g., ['xxx yyy zzz', ...]
+        hyps = bpe_model.decode(token_ids)
+
+        # hyps is a list of list of str, e.g., [['xxx', 'yyy', 'zzz'], ... ]
+        hyps = [s.split() for s in hyps]
+        key = "ctc-decoding"
+        return {key: hyps}
+
+    if params.method == "nbest-oracle":
+        # Note: You can also pass rescored lattices to it.
+        # We choose the HLG decoded lattice for speed reasons
+        # as HLG decoding is faster and the oracle WER
+        # is only slightly worse than that of rescored lattices.
+        best_path = nbest_oracle(
+            lattice=lattice,
+            num_paths=params.num_paths,
+            ref_texts=supervisions["text"],
+            word_table=word_table,
+            nbest_scale=params.nbest_scale,
+            oov="<UNK>",
+        )
+        hyps = get_texts(best_path)
+        hyps = [[word_table[i] for i in ids] for ids in hyps]
+        key = f"oracle_{params.num_paths}_nbest_scale_{params.nbest_scale}"  # noqa
+        return {key: hyps}
+
+    if params.method in ["1best", "nbest"]:
+        if params.method == "1best":
+            best_path = one_best_decoding(
+                lattice=lattice, use_double_scores=params.use_double_scores
+            )
+            key = "no_rescore"
+        else:
+            best_path = nbest_decoding(
+                lattice=lattice,
+                num_paths=params.num_paths,
+                use_double_scores=params.use_double_scores,
+                nbest_scale=params.nbest_scale,
+            )
+            key = f"no_rescore-nbest-scale-{params.nbest_scale}-{params.num_paths}"  # noqa
+
+        hyps = get_texts(best_path)
+        hyps = [[word_table[i] for i in ids] for ids in hyps]
+        return {key: hyps}
+
+    assert params.method in [
+        "nbest-rescoring",
+        "whole-lattice-rescoring",
+        "attention-decoder",
+        "rnn-lm",
+    ]
+
+    lm_scale_list = [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7]
+    lm_scale_list += [0.8, 0.9, 1.0, 1.1, 1.2, 1.3]
+    lm_scale_list += [1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0]
+
+    if params.method == "nbest-rescoring":
+        best_path_dict = rescore_with_n_best_list(
+            lattice=lattice,
+            G=G,
+            num_paths=params.num_paths,
+            lm_scale_list=lm_scale_list,
+            nbest_scale=params.nbest_scale,
+        )
+    elif params.method == "whole-lattice-rescoring":
+        best_path_dict = rescore_with_whole_lattice(
+            lattice=lattice,
+            G_with_epsilon_loops=G,
+            lm_scale_list=lm_scale_list,
+        )
+    elif params.method == "attention-decoder":
+        # lattice uses a 3-gram Lm. We rescore it with a 4-gram LM.
+        rescored_lattice = rescore_with_whole_lattice(
+            lattice=lattice,
+            G_with_epsilon_loops=G,
+            lm_scale_list=None,
+        )
+
+        best_path_dict = rescore_with_attention_decoder(
+            lattice=rescored_lattice,
+            num_paths=params.num_paths,
+            model=model,
+            memory=memory,
+            memory_key_padding_mask=memory_key_padding_mask,
+            sos_id=sos_id,
+            eos_id=eos_id,
+            nbest_scale=params.nbest_scale,
+        )
+    elif params.method == "rnn-lm":
+        # lattice uses a 3-gram Lm. We rescore it with a 4-gram LM.
+        rescored_lattice = rescore_with_whole_lattice(
+            lattice=lattice,
+            G_with_epsilon_loops=G,
+            lm_scale_list=None,
+        )
+
+        best_path_dict = rescore_with_rnn_lm(
+            lattice=rescored_lattice,
+            num_paths=params.num_paths,
+            rnn_lm_model=rnn_lm_model,
+            model=model,
+            memory=memory,
+            memory_key_padding_mask=memory_key_padding_mask,
+            sos_id=sos_id,
+            eos_id=eos_id,
+            blank_id=0,
+            nbest_scale=params.nbest_scale,
+        )
+    else:
+        assert False, f"Unsupported decoding method: {params.method}"
+
+    ans = dict()
+    if best_path_dict is not None:
+        for lm_scale_str, best_path in best_path_dict.items():
+            hyps = get_texts(best_path)
+            hyps = [[word_table[i] for i in ids] for ids in hyps]
+            ans[lm_scale_str] = hyps
+    else:
+        ans = None
+    return ans
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    rnn_lm_model: Optional[nn.Module],
+    HLG: Optional[k2.Fsa],
+    H: Optional[k2.Fsa],
+    bpe_model: Optional[spm.SentencePieceProcessor],
+    word_table: k2.SymbolTable,
+    sos_id: int,
+    eos_id: int,
+    G: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      rnn_lm_model:
+        The neural model for RNN LM.
+      HLG:
+        The decoding graph. Used only when params.method is NOT ctc-decoding.
+      H:
+        The ctc topo. Used only when params.method is ctc-decoding.
+      bpe_model:
+        The BPE model. Used only when params.method is ctc-decoding.
+      word_table:
+        It is the word symbol table.
+      sos_id:
+        The token ID for SOS.
+      eos_id:
+        The token ID for EOS.
+      G:
+        An LM. It is not None when params.method is "nbest-rescoring"
+        or "whole-lattice-rescoring". In general, the G in HLG
+        is a 3-gram LM, while this G is a 4-gram LM.
+    Returns:
+      Return a dict, whose key may be "no-rescore" if no LM rescoring
+      is used, or it may be "lm_scale_0.7" if LM rescoring is used.
+      Its value is a list of tuples. Each tuple contains two elements:
+      The first is the reference transcript, and the second is the
+      predicted result.
+    """
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+        cut_ids = [cut.id for cut in batch["supervisions"]["cut"]]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            rnn_lm_model=rnn_lm_model,
+            HLG=HLG,
+            H=H,
+            bpe_model=bpe_model,
+            batch=batch,
+            word_table=word_table,
+            G=G,
+            sos_id=sos_id,
+            eos_id=eos_id,
+        )
+
+        if hyps_dict is not None:
+            for lm_scale, hyps in hyps_dict.items():
+                this_batch = []
+                assert len(hyps) == len(texts)
+                for cut_id, hyp_words, ref_text in zip(cut_ids, hyps, texts):
+                    ref_words = ref_text.split()
+                    this_batch.append((cut_id, ref_words, hyp_words))
+
+                results[lm_scale].extend(this_batch)
+        else:
+            assert len(results) > 0, "It should not decode to empty in the first batch!"
+            this_batch = []
+            hyp_words = []
+            for ref_text in texts:
+                ref_words = ref_text.split()
+                this_batch.append((ref_words, hyp_words))
+
+            for lm_scale in results.keys():
+                results[lm_scale].extend(this_batch)
+
+        num_cuts += len(texts)
+
+        if batch_idx % 100 == 0:
+            batch_str = f"{batch_idx}/{num_batches}"
+
+            logging.info(f"batch {batch_str}, cuts processed until now is {num_cuts}")
+    return results
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[str, List[int], List[int]]]],
+):
+    if params.method in ("attention-decoder", "rnn-lm"):
+        # Set it to False since there are too many logs.
+        enable_log = False
+    else:
+        enable_log = True
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = params.exp_dir / f"recogs-{test_set_name}-{key}.txt"
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        if enable_log:
+            logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = params.exp_dir / f"errs-{test_set_name}-{key}.txt"
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results, enable_log=enable_log
+            )
+            test_set_wers[key] = wer
+
+        if enable_log:
+            logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = params.exp_dir / f"wer-summary-{test_set_name}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    SwitchBoardAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+    args.lang_dir = Path(args.lang_dir)
+    args.lm_dir = Path(args.lm_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    setup_logger(f"{params.exp_dir}/log-{params.method}/log-decode")
+    logging.info("Decoding started")
+    logging.info(params)
+
+    lexicon = Lexicon(params.lang_dir)
+    max_token_id = max(lexicon.tokens)
+    num_classes = max_token_id + 1  # +1 for the blank
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    graph_compiler = BpeCtcTrainingGraphCompiler(
+        params.lang_dir,
+        device=device,
+        sos_token="<sos/eos>",
+        eos_token="<sos/eos>",
+    )
+    sos_id = graph_compiler.sos_id
+    eos_id = graph_compiler.eos_id
+
+    params.num_classes = num_classes
+    params.sos_id = sos_id
+    params.eos_id = eos_id
+
+    if params.method == "ctc-decoding":
+        HLG = None
+        H = k2.ctc_topo(
+            max_token=max_token_id,
+            modified=False,
+            device=device,
+        )
+        bpe_model = spm.SentencePieceProcessor()
+        bpe_model.load(str(params.lang_dir / "bpe.model"))
+    else:
+        H = None
+        bpe_model = None
+        HLG = k2.Fsa.from_dict(
+            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device)
+        )
+        assert HLG.requires_grad is False
+
+        if not hasattr(HLG, "lm_scores"):
+            HLG.lm_scores = HLG.scores.clone()
+
+    if params.method in (
+        "nbest-rescoring",
+        "whole-lattice-rescoring",
+        "attention-decoder",
+        "rnn-lm",
+    ):
+        if not (params.lm_dir / "G_4_gram.pt").is_file():
+            logging.info("Loading G_4_gram.fst.txt")
+            logging.warning("It may take 8 minutes.")
+            with open(params.lm_dir / "G_4_gram.fst.txt") as f:
+                first_word_disambig_id = lexicon.word_table["#0"]
+
+                G = k2.Fsa.from_openfst(f.read(), acceptor=False)
+                # G.aux_labels is not needed in later computations, so
+                # remove it here.
+                del G.aux_labels
+                # CAUTION: The following line is crucial.
+                # Arcs entering the back-off state have label equal to #0.
+                # We have to change it to 0 here.
+                G.labels[G.labels >= first_word_disambig_id] = 0
+                # See https://github.com/k2-fsa/k2/issues/874
+                # for why we need to set G.properties to None
+                G.__dict__["_properties"] = None
+                G = k2.Fsa.from_fsas([G]).to(device)
+                G = k2.arc_sort(G)
+                # Save a dummy value so that it can be loaded in C++.
+                # See https://github.com/pytorch/pytorch/issues/67902
+                # for why we need to do this.
+                G.dummy = 1
+
+                torch.save(G.as_dict(), params.lm_dir / "G_4_gram.pt")
+        else:
+            logging.info("Loading pre-compiled G_4_gram.pt")
+            d = torch.load(params.lm_dir / "G_4_gram.pt", map_location=device)
+            G = k2.Fsa.from_dict(d)
+
+        if params.method in [
+            "whole-lattice-rescoring",
+            "attention-decoder",
+            "rnn-lm",
+        ]:
+            # Add epsilon self-loops to G as we will compose
+            # it with the whole lattice later
+            G = k2.add_epsilon_self_loops(G)
+            G = k2.arc_sort(G)
+            G = G.to(device)
+
+        # G.lm_scores is used to replace HLG.lm_scores during
+        # LM rescoring.
+        G.lm_scores = G.scores.clone()
+    else:
+        G = None
+
+    model = Conformer(
+        num_features=params.feature_dim,
+        nhead=params.nhead,
+        d_model=params.attention_dim,
+        num_classes=num_classes,
+        subsampling_factor=params.subsampling_factor,
+        num_decoder_layers=params.num_decoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+
+    if params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        model = load_averaged_model(
+            params.exp_dir, model, params.epoch, params.avg, device
+        )
+
+    model.to(device)
+    model.eval()
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    rnn_lm_model = None
+    if params.method == "rnn-lm":
+        rnn_lm_model = RnnLmModel(
+            vocab_size=params.num_classes,
+            embedding_dim=params.rnn_lm_embedding_dim,
+            hidden_dim=params.rnn_lm_hidden_dim,
+            num_layers=params.rnn_lm_num_layers,
+            tie_weights=params.rnn_lm_tie_weights,
+        )
+        if params.rnn_lm_avg == 1:
+            load_checkpoint(
+                f"{params.rnn_lm_exp_dir}/epoch-{params.rnn_lm_epoch}.pt",
+                rnn_lm_model,
+            )
+            rnn_lm_model.to(device)
+        else:
+            rnn_lm_model = load_averaged_model(
+                params.rnn_lm_exp_dir,
+                rnn_lm_model,
+                params.rnn_lm_epoch,
+                params.rnn_lm_avg,
+                device,
+            )
+        rnn_lm_model.eval()
+
+    # we need cut ids to display recognition results.
+    args.return_cuts = True
+    switchboard = SwitchBoardAsrDataModule(args)
+
+    test_eval2000_cuts = switchboard.test_eval2000_cuts().trim_to_supervisions(keep_all_channels=True)
+    test_rt03_cuts = switchboard.test_rt03_cuts().trim_to_supervisions(keep_all_channels=True)
+
+    test_eval2000_dl = switchboard.test_dataloaders(test_eval2000_cuts)
+    test_rt03_dl = switchboard.test_dataloaders(test_rt03_cuts)
+
+    test_sets = ["test-eval2000", "test-rt03"]
+    test_dl = [test_eval2000_dl, test_rt03_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            rnn_lm_model=rnn_lm_model,
+            HLG=HLG,
+            H=H,
+            bpe_model=bpe_model,
+            word_table=lexicon.word_table,
+            G=G,
+            sos_id=sos_id,
+            eos_id=eos_id,
+        )
+
+        save_results(params=params, test_set_name=test_set, results_dict=results_dict)
+
+    logging.info("Done!")
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/conformer_ctc/export.py

@@ -0,0 +1,163 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This script converts several saved checkpoints
+# to a single one using model averaging.
+
+import argparse
+import logging
+from pathlib import Path
+
+import torch
+from conformer import Conformer
+
+from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.lexicon import Lexicon
+from icefall.utils import AttributeDict, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=34,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=20,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="conformer_ctc/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_bpe_500",
+        help="""It contains language related input files such as "lexicon.txt"
+        """,
+    )
+
+    parser.add_argument(
+        "--jit",
+        type=str2bool,
+        default=True,
+        help="""True to save a model after applying torch.jit.script.
+        """,
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            "feature_dim": 80,
+            "subsampling_factor": 4,
+            "use_feat_batchnorm": True,
+            "attention_dim": 512,
+            "nhead": 8,
+            "num_decoder_layers": 6,
+        }
+    )
+    return params
+
+
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+    args.lang_dir = Path(args.lang_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    logging.info(params)
+
+    lexicon = Lexicon(params.lang_dir)
+    max_token_id = max(lexicon.tokens)
+    num_classes = max_token_id + 1  # +1 for the blank
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    model = Conformer(
+        num_features=params.feature_dim,
+        nhead=params.nhead,
+        d_model=params.attention_dim,
+        num_classes=num_classes,
+        subsampling_factor=params.subsampling_factor,
+        num_decoder_layers=params.num_decoder_layers,
+        vgg_frontend=False,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    model.to(device)
+
+    if params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.load_state_dict(average_checkpoints(filenames))
+
+    model.to("cpu")
+    model.eval()
+
+    if params.jit:
+        logging.info("Using torch.jit.script")
+        model = torch.jit.script(model)
+        filename = params.exp_dir / "cpu_jit.pt"
+        model.save(str(filename))
+        logging.info(f"Saved to {filename}")
+    else:
+        logging.info("Not using torch.jit.script")
+        # Save it using a format so that it can be loaded
+        # by :func:`load_checkpoint`
+        filename = params.exp_dir / "pretrained.pt"
+        torch.save({"model": model.state_dict()}, str(filename))
+        logging.info(f"Saved to {filename}")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/swbd/ASR/conformer_ctc/label_smoothing.py

@@ -0,0 +1,109 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+
+
+class LabelSmoothingLoss(torch.nn.Module):
+    """
+    Implement the LabelSmoothingLoss proposed in the following paper
+    https://arxiv.org/pdf/1512.00567.pdf
+    (Rethinking the Inception Architecture for Computer Vision)
+
+    """
+
+    def __init__(
+        self,
+        ignore_index: int = -1,
+        label_smoothing: float = 0.1,
+        reduction: str = "sum",
+    ) -> None:
+        """
+        Args:
+          ignore_index:
+            ignored class id
+          label_smoothing:
+            smoothing rate (0.0 means the conventional cross entropy loss)
+          reduction:
+            It has the same meaning as the reduction in
+            `torch.nn.CrossEntropyLoss`. It can be one of the following three
+            values: (1) "none": No reduction will be applied. (2) "mean": the
+            mean of the output is taken. (3) "sum": the output will be summed.
+        """
+        super().__init__()
+        assert 0.0 <= label_smoothing < 1.0, f"{label_smoothing}"
+        assert reduction in ("none", "sum", "mean"), reduction
+        self.ignore_index = ignore_index
+        self.label_smoothing = label_smoothing
+        self.reduction = reduction
+
+    def forward(self, x: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
+        """
+        Compute loss between x and target.
+
+        Args:
+          x:
+            prediction of dimension
+            (batch_size, input_length, number_of_classes).
+          target:
+            target masked with self.ignore_index of
+            dimension (batch_size, input_length).
+
+        Returns:
+          A scalar tensor containing the loss without normalization.
+        """
+        assert x.ndim == 3
+        assert target.ndim == 2
+        assert x.shape[:2] == target.shape
+        num_classes = x.size(-1)
+        x = x.reshape(-1, num_classes)
+        # Now x is of shape (N*T, C)
+
+        # We don't want to change target in-place below,
+        # so we make a copy of it here
+        target = target.clone().reshape(-1)
+
+        ignored = target == self.ignore_index
+
+        # See https://github.com/k2-fsa/icefall/issues/240
+        # and https://github.com/k2-fsa/icefall/issues/297
+        # for why we don't use target[ignored] = 0 here
+        target = torch.where(ignored, torch.zeros_like(target), target)
+
+        true_dist = torch.nn.functional.one_hot(target, num_classes=num_classes).to(x)
+
+        true_dist = (
+            true_dist * (1 - self.label_smoothing) + self.label_smoothing / num_classes
+        )
+
+        # Set the value of ignored indexes to 0
+        #
+        # See https://github.com/k2-fsa/icefall/issues/240
+        # and https://github.com/k2-fsa/icefall/issues/297
+        # for why we don't use true_dist[ignored] = 0 here
+        true_dist = torch.where(
+            ignored.unsqueeze(1).repeat(1, true_dist.shape[1]),
+            torch.zeros_like(true_dist),
+            true_dist,
+        )
+
+        loss = -1 * (torch.log_softmax(x, dim=1) * true_dist)
+        if self.reduction == "sum":
+            return loss.sum()
+        elif self.reduction == "mean":
+            return loss.sum() / (~ignored).sum()
+        else:
+            return loss.sum(dim=-1)

egs/swbd/ASR/conformer_ctc/pretrained.py

@@ -0,0 +1,430 @@
+#!/usr/bin/env python3
+# Copyright      2021  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                    Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import argparse
+import logging
+import math
+from typing import List
+
+import k2
+import kaldifeat
+import sentencepiece as spm
+import torch
+import torchaudio
+from conformer import Conformer
+from torch.nn.utils.rnn import pad_sequence
+
+from icefall.decode import (
+    get_lattice,
+    one_best_decoding,
+    rescore_with_attention_decoder,
+    rescore_with_whole_lattice,
+)
+from icefall.utils import AttributeDict, get_texts
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--checkpoint",
+        type=str,
+        required=True,
+        help="Path to the checkpoint. "
+        "The checkpoint is assumed to be saved by "
+        "icefall.checkpoint.save_checkpoint().",
+    )
+
+    parser.add_argument(
+        "--words-file",
+        type=str,
+        help="""Path to words.txt.
+        Used only when method is not ctc-decoding.
+        """,
+    )
+
+    parser.add_argument(
+        "--HLG",
+        type=str,
+        help="""Path to HLG.pt.
+        Used only when method is not ctc-decoding.
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        help="""Path to bpe.model.
+        Used only when method is ctc-decoding.
+        """,
+    )
+
+    parser.add_argument(
+        "--method",
+        type=str,
+        default="1best",
+        help="""Decoding method.
+        Possible values are:
+        (0) ctc-decoding - Use CTC decoding. It uses a sentence
+            piece model, i.e., lang_dir/bpe.model, to convert
+            word pieces to words. It needs neither a lexicon
+            nor an n-gram LM.
+        (1) 1best - Use the best path as decoding output. Only
+            the transformer encoder output is used for decoding.
+            We call it HLG decoding.
+        (2) whole-lattice-rescoring - Use an LM to rescore the
+            decoding lattice and then use 1best to decode the
+            rescored lattice.
+            We call it HLG decoding + n-gram LM rescoring.
+        (3) attention-decoder - Extract n paths from the rescored
+            lattice and use the transformer attention decoder for
+            rescoring.
+            We call it HLG decoding + n-gram LM rescoring + attention
+            decoder rescoring.
+        """,
+    )
+
+    parser.add_argument(
+        "--G",
+        type=str,
+        help="""An LM for rescoring.
+        Used only when method is
+        whole-lattice-rescoring or attention-decoder.
+        It's usually a 4-gram LM.
+        """,
+    )
+
+    parser.add_argument(
+        "--num-paths",
+        type=int,
+        default=100,
+        help="""
+        Used only when method is attention-decoder.
+        It specifies the size of n-best list.""",
+    )
+
+    parser.add_argument(
+        "--ngram-lm-scale",
+        type=float,
+        default=1.3,
+        help="""
+        Used only when method is whole-lattice-rescoring and attention-decoder.
+        It specifies the scale for n-gram LM scores.
+        (Note: You need to tune it on a dataset.)
+        """,
+    )
+
+    parser.add_argument(
+        "--attention-decoder-scale",
+        type=float,
+        default=1.2,
+        help="""
+        Used only when method is attention-decoder.
+        It specifies the scale for attention decoder scores.
+        (Note: You need to tune it on a dataset.)
+        """,
+    )
+
+    parser.add_argument(
+        "--nbest-scale",
+        type=float,
+        default=0.5,
+        help="""
+        Used only when method is attention-decoder.
+        It specifies the scale for lattice.scores when
+        extracting n-best lists. A smaller value results in
+        more unique number of paths with the risk of missing
+        the best path.
+        """,
+    )
+
+    parser.add_argument(
+        "--sos-id",
+        type=int,
+        default=1,
+        help="""
+        Used only when method is attention-decoder.
+        It specifies ID for the SOS token.
+        """,
+    )
+
+    parser.add_argument(
+        "--num-classes",
+        type=int,
+        default=500,
+        help="""
+        Vocab size in the BPE model.
+        """,
+    )
+
+    parser.add_argument(
+        "--eos-id",
+        type=int,
+        default=1,
+        help="""
+        Used only when method is attention-decoder.
+        It specifies ID for the EOS token.
+        """,
+    )
+
+    parser.add_argument(
+        "sound_files",
+        type=str,
+        nargs="+",
+        help="The input sound file(s) to transcribe. "
+        "Supported formats are those supported by torchaudio.load(). "
+        "For example, wav and flac are supported. "
+        "The sample rate has to be 16kHz.",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            "sample_rate": 16000,
+            # parameters for conformer
+            "subsampling_factor": 4,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            "feature_dim": 80,
+            "nhead": 8,
+            "attention_dim": 512,
+            "num_decoder_layers": 6,
+            # parameters for decoding
+            "search_beam": 20,
+            "output_beam": 8,
+            "min_active_states": 30,
+            "max_active_states": 10000,
+            "use_double_scores": True,
+        }
+    )
+    return params
+
+
+def read_sound_files(
+    filenames: List[str], expected_sample_rate: float
+) -> List[torch.Tensor]:
+    """Read a list of sound files into a list 1-D float32 torch tensors.
+    Args:
+      filenames:
+        A list of sound filenames.
+      expected_sample_rate:
+        The expected sample rate of the sound files.
+    Returns:
+      Return a list of 1-D float32 torch tensors.
+    """
+    ans = []
+    for f in filenames:
+        wave, sample_rate = torchaudio.load(f)
+        assert (
+            sample_rate == expected_sample_rate
+        ), f"expected sample rate: {expected_sample_rate}. Given: {sample_rate}"
+        # We use only the first channel
+        ans.append(wave[0])
+    return ans
+
+
+def main():
+    parser = get_parser()
+    args = parser.parse_args()
+
+    params = get_params()
+    if args.method != "attention-decoder":
+        # to save memory as the attention decoder
+        # will not be used
+        params.num_decoder_layers = 0
+
+    params.update(vars(args))
+    logging.info(f"{params}")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    logging.info("Creating model")
+    model = Conformer(
+        num_features=params.feature_dim,
+        nhead=params.nhead,
+        d_model=params.attention_dim,
+        num_classes=params.num_classes,
+        subsampling_factor=params.subsampling_factor,
+        num_decoder_layers=params.num_decoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+
+    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    model.load_state_dict(checkpoint["model"], strict=False)
+    model.to(device)
+    model.eval()
+
+    logging.info("Constructing Fbank computer")
+    opts = kaldifeat.FbankOptions()
+    opts.device = device
+    opts.frame_opts.dither = 0
+    opts.frame_opts.snip_edges = False
+    opts.frame_opts.samp_freq = params.sample_rate
+    opts.mel_opts.num_bins = params.feature_dim
+
+    fbank = kaldifeat.Fbank(opts)
+
+    logging.info(f"Reading sound files: {params.sound_files}")
+    waves = read_sound_files(
+        filenames=params.sound_files, expected_sample_rate=params.sample_rate
+    )
+    waves = [w.to(device) for w in waves]
+
+    logging.info("Decoding started")
+    features = fbank(waves)
+
+    features = pad_sequence(features, batch_first=True, padding_value=math.log(1e-10))
+
+    # Note: We don't use key padding mask for attention during decoding
+    with torch.no_grad():
+        nnet_output, memory, memory_key_padding_mask = model(features)
+
+    batch_size = nnet_output.shape[0]
+    supervision_segments = torch.tensor(
+        [[i, 0, nnet_output.shape[1]] for i in range(batch_size)],
+        dtype=torch.int32,
+    )
+
+    if params.method == "ctc-decoding":
+        logging.info("Use CTC decoding")
+        bpe_model = spm.SentencePieceProcessor()
+        bpe_model.load(params.bpe_model)
+        max_token_id = params.num_classes - 1
+
+        H = k2.ctc_topo(
+            max_token=max_token_id,
+            modified=params.num_classes > 500,
+            device=device,
+        )
+
+        lattice = get_lattice(
+            nnet_output=nnet_output,
+            decoding_graph=H,
+            supervision_segments=supervision_segments,
+            search_beam=params.search_beam,
+            output_beam=params.output_beam,
+            min_active_states=params.min_active_states,
+            max_active_states=params.max_active_states,
+            subsampling_factor=params.subsampling_factor,
+        )
+
+        best_path = one_best_decoding(
+            lattice=lattice, use_double_scores=params.use_double_scores
+        )
+        token_ids = get_texts(best_path)
+        hyps = bpe_model.decode(token_ids)
+        hyps = [s.split() for s in hyps]
+    elif params.method in [
+        "1best",
+        "whole-lattice-rescoring",
+        "attention-decoder",
+    ]:
+        logging.info(f"Loading HLG from {params.HLG}")
+        HLG = k2.Fsa.from_dict(torch.load(params.HLG, map_location="cpu"))
+        HLG = HLG.to(device)
+        if not hasattr(HLG, "lm_scores"):
+            # For whole-lattice-rescoring and attention-decoder
+            HLG.lm_scores = HLG.scores.clone()
+
+        if params.method in [
+            "whole-lattice-rescoring",
+            "attention-decoder",
+        ]:
+            logging.info(f"Loading G from {params.G}")
+            G = k2.Fsa.from_dict(torch.load(params.G, map_location="cpu"))
+            # Add epsilon self-loops to G as we will compose
+            # it with the whole lattice later
+            G = G.to(device)
+            G = k2.add_epsilon_self_loops(G)
+            G = k2.arc_sort(G)
+            G.lm_scores = G.scores.clone()
+
+        lattice = get_lattice(
+            nnet_output=nnet_output,
+            decoding_graph=HLG,
+            supervision_segments=supervision_segments,
+            search_beam=params.search_beam,
+            output_beam=params.output_beam,
+            min_active_states=params.min_active_states,
+            max_active_states=params.max_active_states,
+            subsampling_factor=params.subsampling_factor,
+        )
+
+        if params.method == "1best":
+            logging.info("Use HLG decoding")
+            best_path = one_best_decoding(
+                lattice=lattice, use_double_scores=params.use_double_scores
+            )
+        elif params.method == "whole-lattice-rescoring":
+            logging.info("Use HLG decoding + LM rescoring")
+            best_path_dict = rescore_with_whole_lattice(
+                lattice=lattice,
+                G_with_epsilon_loops=G,
+                lm_scale_list=[params.ngram_lm_scale],
+            )
+            best_path = next(iter(best_path_dict.values()))
+        elif params.method == "attention-decoder":
+            logging.info("Use HLG + LM rescoring + attention decoder rescoring")
+            rescored_lattice = rescore_with_whole_lattice(
+                lattice=lattice, G_with_epsilon_loops=G, lm_scale_list=None
+            )
+            best_path_dict = rescore_with_attention_decoder(
+                lattice=rescored_lattice,
+                num_paths=params.num_paths,
+                model=model,
+                memory=memory,
+                memory_key_padding_mask=memory_key_padding_mask,
+                sos_id=params.sos_id,
+                eos_id=params.eos_id,
+                nbest_scale=params.nbest_scale,
+                ngram_lm_scale=params.ngram_lm_scale,
+                attention_scale=params.attention_decoder_scale,
+            )
+            best_path = next(iter(best_path_dict.values()))
+
+        hyps = get_texts(best_path)
+        word_sym_table = k2.SymbolTable.from_file(params.words_file)
+        hyps = [[word_sym_table[i] for i in ids] for ids in hyps]
+    else:
+        raise ValueError(f"Unsupported decoding method: {params.method}")
+
+    s = "\n"
+    for filename, hyp in zip(params.sound_files, hyps):
+        words = " ".join(hyp)
+        s += f"{filename}:\n{words}\n\n"
+    logging.info(s)
+
+    logging.info("Decoding Done")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/swbd/ASR/conformer_ctc/subsampling.py

@@ -0,0 +1,153 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import torch
+import torch.nn as nn
+
+
+class Conv2dSubsampling(nn.Module):
+    """Convolutional 2D subsampling (to 1/4 length).
+
+    Convert an input of shape (N, T, idim) to an output
+    with shape (N, T', odim), where
+    T' = ((T-1)//2 - 1)//2, which approximates T' == T//4
+
+    It is based on
+    https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/subsampling.py  # noqa
+    """
+
+    def __init__(self, idim: int, odim: int) -> None:
+        """
+        Args:
+          idim:
+            Input dim. The input shape is (N, T, idim).
+            Caution: It requires: T >=7, idim >=7
+          odim:
+            Output dim. The output shape is (N, ((T-1)//2 - 1)//2, odim)
+        """
+        assert idim >= 7
+        super().__init__()
+        self.conv = nn.Sequential(
+            nn.Conv2d(in_channels=1, out_channels=odim, kernel_size=3, stride=2),
+            nn.ReLU(),
+            nn.Conv2d(in_channels=odim, out_channels=odim, kernel_size=3, stride=2),
+            nn.ReLU(),
+        )
+        self.out = nn.Linear(odim * (((idim - 1) // 2 - 1) // 2), odim)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """Subsample x.
+
+        Args:
+          x:
+            Its shape is (N, T, idim).
+
+        Returns:
+          Return a tensor of shape (N, ((T-1)//2 - 1)//2, odim)
+        """
+        # On entry, x is (N, T, idim)
+        x = x.unsqueeze(1)  # (N, T, idim) -> (N, 1, T, idim) i.e., (N, C, H, W)
+        x = self.conv(x)
+        # Now x is of shape (N, odim, ((T-1)//2 - 1)//2, ((idim-1)//2 - 1)//2)
+        b, c, t, f = x.size()
+        x = self.out(x.transpose(1, 2).contiguous().view(b, t, c * f))
+        # Now x is of shape (N, ((T-1)//2 - 1))//2, odim)
+        return x
+
+
+class VggSubsampling(nn.Module):
+    """Trying to follow the setup described in the following paper:
+    https://arxiv.org/pdf/1910.09799.pdf
+
+    This paper is not 100% explicit so I am guessing to some extent,
+    and trying to compare with other VGG implementations.
+
+    Convert an input of shape (N, T, idim) to an output
+    with shape (N, T', odim), where
+    T' = ((T-1)//2 - 1)//2, which approximates T' = T//4
+    """
+
+    def __init__(self, idim: int, odim: int) -> None:
+        """Construct a VggSubsampling object.
+
+        This uses 2 VGG blocks with 2 Conv2d layers each,
+        subsampling its input by a factor of 4 in the time dimensions.
+
+        Args:
+          idim:
+            Input dim. The input shape is (N, T, idim).
+            Caution: It requires: T >=7, idim >=7
+          odim:
+            Output dim. The output shape is (N, ((T-1)//2 - 1)//2, odim)
+        """
+        super().__init__()
+
+        cur_channels = 1
+        layers = []
+        block_dims = [32, 64]
+
+        # The decision to use padding=1 for the 1st convolution, then padding=0
+        # for the 2nd and for the max-pooling, and ceil_mode=True, was driven by
+        # a back-compatibility concern so that the number of frames at the
+        # output would be equal to:
+        #  (((T-1)//2)-1)//2.
+        # We can consider changing this by using padding=1 on the
+        # 2nd convolution, so the num-frames at the output would be T//4.
+        for block_dim in block_dims:
+            layers.append(
+                torch.nn.Conv2d(
+                    in_channels=cur_channels,
+                    out_channels=block_dim,
+                    kernel_size=3,
+                    padding=1,
+                    stride=1,
+                )
+            )
+            layers.append(torch.nn.ReLU())
+            layers.append(
+                torch.nn.Conv2d(
+                    in_channels=block_dim,
+                    out_channels=block_dim,
+                    kernel_size=3,
+                    padding=0,
+                    stride=1,
+                )
+            )
+            layers.append(
+                torch.nn.MaxPool2d(kernel_size=2, stride=2, padding=0, ceil_mode=True)
+            )
+            cur_channels = block_dim
+
+        self.layers = nn.Sequential(*layers)
+
+        self.out = nn.Linear(block_dims[-1] * (((idim - 1) // 2 - 1) // 2), odim)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """Subsample x.
+
+        Args:
+          x:
+            Its shape is (N, T, idim).
+
+        Returns:
+          Return a tensor of shape (N, ((T-1)//2 - 1)//2, odim)
+        """
+        x = x.unsqueeze(1)
+        x = self.layers(x)
+        b, c, t, f = x.size()
+        x = self.out(x.transpose(1, 2).contiguous().view(b, t, c * f))
+        return x

egs/swbd/ASR/conformer_ctc/test_label_smoothing.py

@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from distutils.version import LooseVersion
+
+import torch
+from label_smoothing import LabelSmoothingLoss
+
+torch_ver = LooseVersion(torch.__version__)
+
+
+def test_with_torch_label_smoothing_loss():
+    if torch_ver < LooseVersion("1.10.0"):
+        print(f"Current torch version: {torch_ver}")
+        print("Please use torch >= 1.10 to run this test - skipping")
+        return
+    torch.manual_seed(20211105)
+    x = torch.rand(20, 30, 5000)
+    tgt = torch.randint(low=-1, high=x.size(-1), size=x.shape[:2])
+    for reduction in ["none", "sum", "mean"]:
+        custom_loss_func = LabelSmoothingLoss(
+            ignore_index=-1, label_smoothing=0.1, reduction=reduction
+        )
+        custom_loss = custom_loss_func(x, tgt)
+
+        torch_loss_func = torch.nn.CrossEntropyLoss(
+            ignore_index=-1, reduction=reduction, label_smoothing=0.1
+        )
+        torch_loss = torch_loss_func(x.reshape(-1, x.size(-1)), tgt.reshape(-1))
+        assert torch.allclose(custom_loss, torch_loss)
+
+
+def main():
+    test_with_torch_label_smoothing_loss()
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/conformer_ctc/test_subsampling.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import torch
+from subsampling import Conv2dSubsampling, VggSubsampling
+
+
+def test_conv2d_subsampling():
+    N = 3
+    odim = 2
+
+    for T in range(7, 19):
+        for idim in range(7, 20):
+            model = Conv2dSubsampling(idim=idim, odim=odim)
+            x = torch.empty(N, T, idim)
+            y = model(x)
+            assert y.shape[0] == N
+            assert y.shape[1] == ((T - 1) // 2 - 1) // 2
+            assert y.shape[2] == odim
+
+
+def test_vgg_subsampling():
+    N = 3
+    odim = 2
+
+    for T in range(7, 19):
+        for idim in range(7, 20):
+            model = VggSubsampling(idim=idim, odim=odim)
+            x = torch.empty(N, T, idim)
+            y = model(x)
+            assert y.shape[0] == N
+            assert y.shape[1] == ((T - 1) // 2 - 1) // 2
+            assert y.shape[2] == odim

egs/swbd/ASR/conformer_ctc/test_transformer.py

@@ -0,0 +1,104 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import torch
+from torch.nn.utils.rnn import pad_sequence
+from transformer import (
+    Transformer,
+    add_eos,
+    add_sos,
+    decoder_padding_mask,
+    encoder_padding_mask,
+    generate_square_subsequent_mask,
+)
+
+
+def test_encoder_padding_mask():
+    supervisions = {
+        "sequence_idx": torch.tensor([0, 1, 2]),
+        "start_frame": torch.tensor([0, 0, 0]),
+        "num_frames": torch.tensor([18, 7, 13]),
+    }
+
+    max_len = ((18 - 1) // 2 - 1) // 2
+    mask = encoder_padding_mask(max_len, supervisions)
+    expected_mask = torch.tensor(
+        [
+            [False, False, False],  # ((18 - 1)//2 - 1)//2 = 3,
+            [False, True, True],  # ((7 - 1)//2 - 1)//2 = 1,
+            [False, False, True],  # ((13 - 1)//2 - 1)//2 = 2,
+        ]
+    )
+    assert torch.all(torch.eq(mask, expected_mask))
+
+
+def test_transformer():
+    num_features = 40
+    num_classes = 87
+    model = Transformer(num_features=num_features, num_classes=num_classes)
+
+    N = 31
+
+    for T in range(7, 30):
+        x = torch.rand(N, T, num_features)
+        y, _, _ = model(x)
+        assert y.shape == (N, (((T - 1) // 2) - 1) // 2, num_classes)
+
+
+def test_generate_square_subsequent_mask():
+    s = 5
+    mask = generate_square_subsequent_mask(s)
+    inf = float("inf")
+    expected_mask = torch.tensor(
+        [
+            [0.0, -inf, -inf, -inf, -inf],
+            [0.0, 0.0, -inf, -inf, -inf],
+            [0.0, 0.0, 0.0, -inf, -inf],
+            [0.0, 0.0, 0.0, 0.0, -inf],
+            [0.0, 0.0, 0.0, 0.0, 0.0],
+        ]
+    )
+    assert torch.all(torch.eq(mask, expected_mask))
+
+
+def test_decoder_padding_mask():
+    x = [torch.tensor([1, 2]), torch.tensor([3]), torch.tensor([2, 5, 8])]
+    y = pad_sequence(x, batch_first=True, padding_value=-1)
+    mask = decoder_padding_mask(y, ignore_id=-1)
+    expected_mask = torch.tensor(
+        [
+            [False, False, True],
+            [False, True, True],
+            [False, False, False],
+        ]
+    )
+    assert torch.all(torch.eq(mask, expected_mask))
+
+
+def test_add_sos():
+    x = [[1, 2], [3], [2, 5, 8]]
+    y = add_sos(x, sos_id=0)
+    expected_y = [[0, 1, 2], [0, 3], [0, 2, 5, 8]]
+    assert y == expected_y
+
+
+def test_add_eos():
+    x = [[1, 2], [3], [2, 5, 8]]
+    y = add_eos(x, eos_id=0)
+    expected_y = [[1, 2, 0], [3, 0], [2, 5, 8, 0]]
+    assert y == expected_y

egs/swbd/ASR/conformer_ctc/train.py

@@ -0,0 +1,813 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                  Wei Kang
+#                                                  Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Usage:
+  export CUDA_VISIBLE_DEVICES="0,1,2,3"
+  ./conformer_ctc/train.py \
+     --exp-dir ./conformer_ctc/exp \
+     --world-size 4 \
+     --max-duration 200 \
+     --num-epochs 20
+"""
+
+import argparse
+import logging
+from pathlib import Path
+from shutil import copyfile
+from typing import Optional, Tuple
+
+import k2
+import torch
+import torch.multiprocessing as mp
+import torch.nn as nn
+from asr_datamodule import SwitchBoardAsrDataModule
+from conformer import Conformer
+from lhotse.cut import Cut
+from lhotse.utils import fix_random_seed
+from torch import Tensor
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.nn.utils import clip_grad_norm_
+from torch.utils.tensorboard import SummaryWriter
+from transformer import Noam
+
+from icefall.bpe_graph_compiler import BpeCtcTrainingGraphCompiler
+from icefall.checkpoint import load_checkpoint
+from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.dist import cleanup_dist, setup_dist
+from icefall.env import get_env_info
+from icefall.graph_compiler import CtcTrainingGraphCompiler
+from icefall.lexicon import Lexicon
+from icefall.utils import (
+    AttributeDict,
+    MetricsTracker,
+    encode_supervisions,
+    setup_logger,
+    str2bool,
+)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--world-size",
+        type=int,
+        default=1,
+        help="Number of GPUs for DDP training.",
+    )
+
+    parser.add_argument(
+        "--master-port",
+        type=int,
+        default=12354,
+        help="Master port to use for DDP training.",
+    )
+
+    parser.add_argument(
+        "--tensorboard",
+        type=str2bool,
+        default=True,
+        help="Should various information be logged in tensorboard.",
+    )
+
+    parser.add_argument(
+        "--num-epochs",
+        type=int,
+        default=78,
+        help="Number of epochs to train.",
+    )
+
+    parser.add_argument(
+        "--start-epoch",
+        type=int,
+        default=0,
+        help="""Resume training from from this epoch.
+        If it is positive, it will load checkpoint from
+        conformer_ctc/exp/epoch-{start_epoch-1}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="conformer_ctc/exp",
+        help="""The experiment dir.
+        It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_bpe_500",
+        help="""The lang dir
+        It contains language related input files such as
+        "lexicon.txt"
+        """,
+    )
+
+    parser.add_argument(
+        "--att-rate",
+        type=float,
+        default=0.8,
+        help="""The attention rate.
+        The total loss is (1 -  att_rate) * ctc_loss + att_rate * att_loss
+        """,
+    )
+
+    parser.add_argument(
+        "--num-decoder-layers",
+        type=int,
+        default=6,
+        help="""Number of decoder layer of transformer decoder.
+        Setting this to 0 will not create the decoder at all (pure CTC model)
+        """,
+    )
+
+    parser.add_argument(
+        "--lr-factor",
+        type=float,
+        default=5.0,
+        help="The lr_factor for Noam optimizer",
+    )
+
+    parser.add_argument(
+        "--seed",
+        type=int,
+        default=42,
+        help="The seed for random generators intended for reproducibility",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    """Return a dict containing training parameters.
+
+    All training related parameters that are not passed from the commandline
+    are saved in the variable `params`.
+
+    Commandline options are merged into `params` after they are parsed, so
+    you can also access them via `params`.
+
+    Explanation of options saved in `params`:
+
+        - best_train_loss: Best training loss so far. It is used to select
+                           the model that has the lowest training loss. It is
+                           updated during the training.
+
+        - best_valid_loss: Best validation loss so far. It is used to select
+                           the model that has the lowest validation loss. It is
+                           updated during the training.
+
+        - best_train_epoch: It is the epoch that has the best training loss.
+
+        - best_valid_epoch: It is the epoch that has the best validation loss.
+
+        - batch_idx_train: Used to writing statistics to tensorboard. It
+                           contains number of batches trained so far across
+                           epochs.
+
+        - log_interval:  Print training loss if batch_idx % log_interval` is 0
+
+        - reset_interval: Reset statistics if batch_idx % reset_interval is 0
+
+        - valid_interval:  Run validation if batch_idx % valid_interval is 0
+
+        - feature_dim: The model input dim. It has to match the one used
+                       in computing features.
+
+        - subsampling_factor:  The subsampling factor for the model.
+
+        - use_feat_batchnorm: Normalization for the input features, can be a
+                              boolean indicating whether to do batch
+                              normalization, or a float which means just scaling
+                              the input features with this float value.
+                              If given a float value, we will remove batchnorm
+                              layer in `ConvolutionModule` as well.
+
+        - attention_dim: Hidden dim for multi-head attention model.
+
+        - head: Number of heads of multi-head attention model.
+
+        - num_decoder_layers: Number of decoder layer of transformer decoder.
+
+        - beam_size: It is used in k2.ctc_loss
+
+        - reduction: It is used in k2.ctc_loss
+
+        - use_double_scores: It is used in k2.ctc_loss
+
+        - weight_decay:  The weight_decay for the optimizer.
+
+        - warm_step: The warm_step for Noam optimizer.
+    """
+    params = AttributeDict(
+        {
+            "best_train_loss": float("inf"),
+            "best_valid_loss": float("inf"),
+            "best_train_epoch": -1,
+            "best_valid_epoch": -1,
+            "batch_idx_train": 0,
+            "log_interval": 50,
+            "reset_interval": 200,
+            "valid_interval": 3000,
+            # parameters for conformer
+            "feature_dim": 80,
+            "subsampling_factor": 4,
+            "use_feat_batchnorm": True,
+            "attention_dim": 512,
+            "nhead": 8,
+            # parameters for loss
+            "beam_size": 10,
+            "reduction": "sum",
+            "use_double_scores": True,
+            # parameters for Noam
+            "weight_decay": 1e-6,
+            "warm_step": 80000,
+            "env_info": get_env_info(),
+        }
+    )
+
+    return params
+
+
+def load_checkpoint_if_available(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+) -> None:
+    """Load checkpoint from file.
+
+    If params.start_epoch is positive, it will load the checkpoint from
+    `params.start_epoch - 1`. Otherwise, this function does nothing.
+
+    Apart from loading state dict for `model`, `optimizer` and `scheduler`,
+    it also updates `best_train_epoch`, `best_train_loss`, `best_valid_epoch`,
+    and `best_valid_loss` in `params`.
+
+    Args:
+      params:
+        The return value of :func:`get_params`.
+      model:
+        The training model.
+      optimizer:
+        The optimizer that we are using.
+      scheduler:
+        The learning rate scheduler we are using.
+    Returns:
+      Return None.
+    """
+    if params.start_epoch <= 0:
+        return
+
+    filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
+    saved_params = load_checkpoint(
+        filename,
+        model=model,
+        optimizer=optimizer,
+        scheduler=scheduler,
+    )
+
+    keys = [
+        "best_train_epoch",
+        "best_valid_epoch",
+        "batch_idx_train",
+        "best_train_loss",
+        "best_valid_loss",
+    ]
+    for k in keys:
+        params[k] = saved_params[k]
+
+    return saved_params
+
+
+def save_checkpoint(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+    rank: int = 0,
+) -> None:
+    """Save model, optimizer, scheduler and training stats to file.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The training model.
+    """
+    if rank != 0:
+        return
+    filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt"
+    save_checkpoint_impl(
+        filename=filename,
+        model=model,
+        params=params,
+        optimizer=optimizer,
+        scheduler=scheduler,
+        rank=rank,
+    )
+
+    if params.best_train_epoch == params.cur_epoch:
+        best_train_filename = params.exp_dir / "best-train-loss.pt"
+        copyfile(src=filename, dst=best_train_filename)
+
+    if params.best_valid_epoch == params.cur_epoch:
+        best_valid_filename = params.exp_dir / "best-valid-loss.pt"
+        copyfile(src=filename, dst=best_valid_filename)
+
+
+def compute_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    batch: dict,
+    graph_compiler: BpeCtcTrainingGraphCompiler,
+    is_training: bool,
+) -> Tuple[Tensor, MetricsTracker]:
+    """
+    Compute CTC loss given the model and its inputs.
+
+    Args:
+      params:
+        Parameters for training. See :func:`get_params`.
+      model:
+        The model for training. It is an instance of Conformer in our case.
+      batch:
+        A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+        for the content in it.
+      graph_compiler:
+        It is used to build a decoding graph from a ctc topo and training
+        transcript. The training transcript is contained in the given `batch`,
+        while the ctc topo is built when this compiler is instantiated.
+      is_training:
+        True for training. False for validation. When it is True, this
+        function enables autograd during computation; when it is False, it
+        disables autograd.
+    """
+    device = graph_compiler.device
+    feature = batch["inputs"]
+    # at entry, feature is (N, T, C)
+    assert feature.ndim == 3
+    feature = feature.to(device)
+
+    supervisions = batch["supervisions"]
+    with torch.set_grad_enabled(is_training):
+        nnet_output, encoder_memory, memory_mask = model(feature, supervisions)
+        # nnet_output is (N, T, C)
+
+    # NOTE: We need `encode_supervisions` to sort sequences with
+    # different duration in decreasing order, required by
+    # `k2.intersect_dense` called in `k2.ctc_loss`
+    supervision_segments, texts = encode_supervisions(
+        supervisions, subsampling_factor=params.subsampling_factor
+    )
+
+    if isinstance(graph_compiler, BpeCtcTrainingGraphCompiler):
+        # Works with a BPE model
+        token_ids = graph_compiler.texts_to_ids(texts)
+        decoding_graph = graph_compiler.compile(token_ids)
+    elif isinstance(graph_compiler, CtcTrainingGraphCompiler):
+        # Works with a phone lexicon
+        decoding_graph = graph_compiler.compile(texts)
+    else:
+        raise ValueError(f"Unsupported type of graph compiler: {type(graph_compiler)}")
+
+    dense_fsa_vec = k2.DenseFsaVec(
+        nnet_output,
+        supervision_segments,
+        allow_truncate=params.subsampling_factor - 1,
+    )
+
+    ctc_loss = k2.ctc_loss(
+        decoding_graph=decoding_graph,
+        dense_fsa_vec=dense_fsa_vec,
+        output_beam=params.beam_size,
+        reduction=params.reduction,
+        use_double_scores=params.use_double_scores,
+    )
+
+    if params.att_rate != 0.0:
+        with torch.set_grad_enabled(is_training):
+            mmodel = model.module if hasattr(model, "module") else model
+            # Note: We need to generate an unsorted version of token_ids
+            # `encode_supervisions()` called above sorts text, but
+            # encoder_memory and memory_mask are not sorted, so we
+            # use an unsorted version `supervisions["text"]` to regenerate
+            # the token_ids
+            #
+            # See https://github.com/k2-fsa/icefall/issues/97
+            # for more details
+            unsorted_token_ids = graph_compiler.texts_to_ids(supervisions["text"])
+            att_loss = mmodel.decoder_forward(
+                encoder_memory,
+                memory_mask,
+                token_ids=unsorted_token_ids,
+                sos_id=graph_compiler.sos_id,
+                eos_id=graph_compiler.eos_id,
+            )
+        loss = (1.0 - params.att_rate) * ctc_loss + params.att_rate * att_loss
+    else:
+        loss = ctc_loss
+        att_loss = torch.tensor([0])
+
+    assert loss.requires_grad == is_training
+
+    info = MetricsTracker()
+    info["frames"] = supervision_segments[:, 2].sum().item()
+    info["ctc_loss"] = ctc_loss.detach().cpu().item()
+    if params.att_rate != 0.0:
+        info["att_loss"] = att_loss.detach().cpu().item()
+
+    info["loss"] = loss.detach().cpu().item()
+
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utterances"] = feature.size(0)
+    # averaged input duration in frames over utterances
+    info["utt_duration"] = supervisions["num_frames"].sum().item()
+    # averaged padding proportion over utterances
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - supervisions["num_frames"]) / feature.size(1)).sum().item()
+    )
+
+    return loss, info
+
+
+def compute_validation_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    graph_compiler: BpeCtcTrainingGraphCompiler,
+    valid_dl: torch.utils.data.DataLoader,
+    world_size: int = 1,
+) -> MetricsTracker:
+    """Run the validation process."""
+    model.eval()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(valid_dl):
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            batch=batch,
+            graph_compiler=graph_compiler,
+            is_training=False,
+        )
+        assert loss.requires_grad is False
+        tot_loss = tot_loss + loss_info
+
+    if world_size > 1:
+        tot_loss.reduce(loss.device)
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    if loss_value < params.best_valid_loss:
+        params.best_valid_epoch = params.cur_epoch
+        params.best_valid_loss = loss_value
+
+    return tot_loss
+
+
+def train_one_epoch(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: torch.optim.Optimizer,
+    graph_compiler: BpeCtcTrainingGraphCompiler,
+    train_dl: torch.utils.data.DataLoader,
+    valid_dl: torch.utils.data.DataLoader,
+    tb_writer: Optional[SummaryWriter] = None,
+    world_size: int = 1,
+) -> None:
+    """Train the model for one epoch.
+
+    The training loss from the mean of all frames is saved in
+    `params.train_loss`. It runs the validation process every
+    `params.valid_interval` batches.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The model for training.
+      optimizer:
+        The optimizer we are using.
+      graph_compiler:
+        It is used to convert transcripts to FSAs.
+      train_dl:
+        Dataloader for the training dataset.
+      valid_dl:
+        Dataloader for the validation dataset.
+      tb_writer:
+        Writer to write log messages to tensorboard.
+      world_size:
+        Number of nodes in DDP training. If it is 1, DDP is disabled.
+    """
+    model.train()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(train_dl):
+        params.batch_idx_train += 1
+        batch_size = len(batch["supervisions"]["text"])
+
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            batch=batch,
+            graph_compiler=graph_compiler,
+            is_training=True,
+        )
+        # summary stats
+        tot_loss = (tot_loss * (1 - 1 / params.reset_interval)) + loss_info
+
+        # NOTE: We use reduction==sum and loss is computed over utterances
+        # in the batch and there is no normalization to it so far.
+
+        optimizer.zero_grad()
+        loss.backward()
+        clip_grad_norm_(model.parameters(), 5.0, 2.0)
+        optimizer.step()
+
+        if batch_idx % params.log_interval == 0:
+            logging.info(
+                f"Epoch {params.cur_epoch}, "
+                f"batch {batch_idx}, loss[{loss_info}], "
+                f"tot_loss[{tot_loss}], batch size: {batch_size}"
+            )
+
+        if batch_idx % params.log_interval == 0:
+            if tb_writer is not None:
+                loss_info.write_summary(
+                    tb_writer, "train/current_", params.batch_idx_train
+                )
+                tot_loss.write_summary(tb_writer, "train/tot_", params.batch_idx_train)
+
+        if batch_idx > 0 and batch_idx % params.valid_interval == 0:
+            logging.info("Computing validation loss")
+            valid_info = compute_validation_loss(
+                params=params,
+                model=model,
+                graph_compiler=graph_compiler,
+                valid_dl=valid_dl,
+                world_size=world_size,
+            )
+            model.train()
+            logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
+            if tb_writer is not None:
+                valid_info.write_summary(
+                    tb_writer, "train/valid_", params.batch_idx_train
+                )
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    params.train_loss = loss_value
+    if params.train_loss < params.best_train_loss:
+        params.best_train_epoch = params.cur_epoch
+        params.best_train_loss = params.train_loss
+
+
+def run(rank, world_size, args):
+    """
+    Args:
+      rank:
+        It is a value between 0 and `world_size-1`, which is
+        passed automatically by `mp.spawn()` in :func:`main`.
+        The node with rank 0 is responsible for saving checkpoint.
+      world_size:
+        Number of GPUs for DDP training.
+      args:
+        The return value of get_parser().parse_args()
+    """
+    params = get_params()
+    params.update(vars(args))
+
+    fix_random_seed(params.seed)
+    if world_size > 1:
+        setup_dist(rank, world_size, params.master_port)
+
+    setup_logger(f"{params.exp_dir}/log/log-train")
+    logging.info("Training started")
+    logging.info(params)
+
+    if args.tensorboard and rank == 0:
+        tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard")
+    else:
+        tb_writer = None
+
+    lexicon = Lexicon(params.lang_dir)
+    max_token_id = max(lexicon.tokens)
+    num_classes = max_token_id + 1  # +1 for the blank
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", rank)
+
+    if "lang_bpe" in str(params.lang_dir):
+        graph_compiler = BpeCtcTrainingGraphCompiler(
+            params.lang_dir,
+            device=device,
+            sos_token="<sos/eos>",
+            eos_token="<sos/eos>",
+        )
+    elif "lang_phone" in str(params.lang_dir):
+        assert params.att_rate == 0, (
+            "Attention decoder training does not support phone lang dirs "
+            "at this time due to a missing <sos/eos> symbol. Set --att-rate=0 "
+            "for pure CTC training when using a phone-based lang dir."
+        )
+        assert params.num_decoder_layers == 0, (
+            "Attention decoder training does not support phone lang dirs "
+            "at this time due to a missing <sos/eos> symbol. "
+            "Set --num-decoder-layers=0 for pure CTC training when using "
+            "a phone-based lang dir."
+        )
+        graph_compiler = CtcTrainingGraphCompiler(
+            lexicon,
+            device=device,
+        )
+        # Manually add the sos/eos ID with their default values
+        # from the BPE recipe which we're adapting here.
+        graph_compiler.sos_id = 1
+        graph_compiler.eos_id = 1
+    else:
+        raise ValueError(
+            f"Unsupported type of lang dir (we expected it to have "
+            f"'lang_bpe' or 'lang_phone' in its name): {params.lang_dir}"
+        )
+
+    logging.info("About to create model")
+    model = Conformer(
+        num_features=params.feature_dim,
+        nhead=params.nhead,
+        d_model=params.attention_dim,
+        num_classes=num_classes,
+        subsampling_factor=params.subsampling_factor,
+        num_decoder_layers=params.num_decoder_layers,
+        vgg_frontend=False,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+
+    checkpoints = load_checkpoint_if_available(params=params, model=model)
+
+    model.to(device)
+    if world_size > 1:
+        model = DDP(model, device_ids=[rank])
+
+    optimizer = Noam(
+        model.parameters(),
+        model_size=params.attention_dim,
+        factor=params.lr_factor,
+        warm_step=params.warm_step,
+        weight_decay=params.weight_decay,
+    )
+
+    if checkpoints:
+        optimizer.load_state_dict(checkpoints["optimizer"])
+
+    switchboard = SwitchBoardAsrDataModule(args)
+
+    train_cuts = switchboard.train_all_cuts()
+
+    def remove_short_and_long_utt(c: Cut):
+        # Keep only utterances with duration between 1 second and 20 seconds
+        #
+        # Caution: There is a reason to select 20.0 here. Please see
+        # ../local/display_manifest_statistics.py
+        #
+        # You should use ../local/display_manifest_statistics.py to get
+        # an utterance duration distribution for your dataset to select
+        # the threshold
+        return 1.0 <= c.duration <= 600.0
+
+    train_cuts = train_cuts.filter(remove_short_and_long_utt)
+
+    train_dl = switchboard.train_dataloaders(train_cuts)
+
+    valid_cuts = switchboard.dev_cuts()
+    valid_dl = switchboard.valid_dataloaders(valid_cuts)
+
+    scan_pessimistic_batches_for_oom(
+        model=model,
+        train_dl=train_dl,
+        optimizer=optimizer,
+        graph_compiler=graph_compiler,
+        params=params,
+    )
+
+    for epoch in range(params.start_epoch, params.num_epochs):
+        fix_random_seed(params.seed + epoch)
+        train_dl.sampler.set_epoch(epoch)
+
+        cur_lr = optimizer._rate
+        if tb_writer is not None:
+            tb_writer.add_scalar("train/learning_rate", cur_lr, params.batch_idx_train)
+            tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
+
+        if rank == 0:
+            logging.info("epoch {}, learning rate {}".format(epoch, cur_lr))
+
+        params.cur_epoch = epoch
+
+        train_one_epoch(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            graph_compiler=graph_compiler,
+            train_dl=train_dl,
+            valid_dl=valid_dl,
+            tb_writer=tb_writer,
+            world_size=world_size,
+        )
+
+        save_checkpoint(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            rank=rank,
+        )
+
+    logging.info("Done!")
+
+    if world_size > 1:
+        torch.distributed.barrier()
+        cleanup_dist()
+
+
+def scan_pessimistic_batches_for_oom(
+    model: nn.Module,
+    train_dl: torch.utils.data.DataLoader,
+    optimizer: torch.optim.Optimizer,
+    graph_compiler: BpeCtcTrainingGraphCompiler,
+    params: AttributeDict,
+):
+    from lhotse.dataset import find_pessimistic_batches
+
+    logging.info(
+        "Sanity check -- see if any of the batches in epoch 0 would cause OOM."
+    )
+    batches, crit_values = find_pessimistic_batches(train_dl.sampler)
+    for criterion, cuts in batches.items():
+        batch = train_dl.dataset[cuts]
+        try:
+            optimizer.zero_grad()
+            loss, _ = compute_loss(
+                params=params,
+                model=model,
+                batch=batch,
+                graph_compiler=graph_compiler,
+                is_training=True,
+            )
+            loss.backward()
+            clip_grad_norm_(model.parameters(), 5.0, 2.0)
+            optimizer.step()
+        except RuntimeError as e:
+            if "CUDA out of memory" in str(e):
+                logging.error(
+                    "Your GPU ran out of memory with the current "
+                    "max_duration setting. We recommend decreasing "
+                    "max_duration and trying again.\n"
+                    f"Failing criterion: {criterion} "
+                    f"(={crit_values[criterion]}) ..."
+                )
+            raise
+
+
+def main():
+    parser = get_parser()
+    SwitchBoardAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+    args.lang_dir = Path(args.lang_dir)
+
+    world_size = args.world_size
+    assert world_size >= 1
+    if world_size > 1:
+        mp.spawn(run, args=(world_size, args), nprocs=world_size, join=True)
+    else:
+        run(rank=0, world_size=1, args=args)
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/conformer_ctc/transformer.py

@@ -0,0 +1,928 @@
+# Copyright    2021 University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import math
+from typing import Dict, List, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+from label_smoothing import LabelSmoothingLoss
+from subsampling import Conv2dSubsampling, VggSubsampling
+from torch.nn.utils.rnn import pad_sequence
+
+# Note: TorchScript requires Dict/List/etc. to be fully typed.
+Supervisions = Dict[str, torch.Tensor]
+
+
+class Transformer(nn.Module):
+    def __init__(
+        self,
+        num_features: int,
+        num_classes: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        num_decoder_layers: int = 6,
+        dropout: float = 0.1,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: Union[float, bool] = 0.1,
+    ) -> None:
+        """
+        Args:
+          num_features:
+            The input dimension of the model.
+          num_classes:
+            The output dimension of the model.
+          subsampling_factor:
+            Number of output frames is num_in_frames // subsampling_factor.
+            Currently, subsampling_factor MUST be 4.
+          d_model:
+            Attention dimension.
+          nhead:
+            Number of heads in multi-head attention.
+            Must satisfy d_model // nhead == 0.
+          dim_feedforward:
+            The output dimension of the feedforward layers in encoder/decoder.
+          num_encoder_layers:
+            Number of encoder layers.
+          num_decoder_layers:
+            Number of decoder layers.
+          dropout:
+            Dropout in encoder/decoder.
+          normalize_before:
+            If True, use pre-layer norm; False to use post-layer norm.
+          vgg_frontend:
+            True to use vgg style frontend for subsampling.
+          use_feat_batchnorm:
+            True to use batchnorm for the input layer.
+            Float value to scale the input layer.
+            False to do nothing.
+        """
+        super().__init__()
+        self.use_feat_batchnorm = use_feat_batchnorm
+        assert isinstance(use_feat_batchnorm, (float, bool))
+        if isinstance(use_feat_batchnorm, bool) and use_feat_batchnorm:
+            self.feat_batchnorm = nn.BatchNorm1d(num_features)
+
+        self.num_features = num_features
+        self.num_classes = num_classes
+        self.subsampling_factor = subsampling_factor
+        if subsampling_factor != 4:
+            raise NotImplementedError("Support only 'subsampling_factor=4'.")
+
+        # self.encoder_embed converts the input of shape (N, T, num_classes)
+        # to the shape (N, T//subsampling_factor, d_model).
+        # That is, it does two things simultaneously:
+        #   (1) subsampling: T -> T//subsampling_factor
+        #   (2) embedding: num_classes -> d_model
+        if vgg_frontend:
+            self.encoder_embed = VggSubsampling(num_features, d_model)
+        else:
+            self.encoder_embed = Conv2dSubsampling(num_features, d_model)
+
+        self.encoder_pos = PositionalEncoding(d_model, dropout)
+
+        encoder_layer = TransformerEncoderLayer(
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            dropout=dropout,
+            normalize_before=normalize_before,
+        )
+
+        if normalize_before:
+            encoder_norm = nn.LayerNorm(d_model)
+        else:
+            encoder_norm = None
+
+        self.encoder = nn.TransformerEncoder(
+            encoder_layer=encoder_layer,
+            num_layers=num_encoder_layers,
+            norm=encoder_norm,
+        )
+
+        # TODO(fangjun): remove dropout
+        self.encoder_output_layer = nn.Sequential(
+            nn.Dropout(p=dropout), nn.Linear(d_model, num_classes)
+        )
+
+        if num_decoder_layers > 0:
+            self.decoder_num_class = (
+                self.num_classes
+            )  # bpe model already has sos/eos symbol
+
+            self.decoder_embed = nn.Embedding(
+                num_embeddings=self.decoder_num_class, embedding_dim=d_model
+            )
+            self.decoder_pos = PositionalEncoding(d_model, dropout)
+
+            decoder_layer = TransformerDecoderLayer(
+                d_model=d_model,
+                nhead=nhead,
+                dim_feedforward=dim_feedforward,
+                dropout=dropout,
+                normalize_before=normalize_before,
+            )
+
+            if normalize_before:
+                decoder_norm = nn.LayerNorm(d_model)
+            else:
+                decoder_norm = None
+
+            self.decoder = nn.TransformerDecoder(
+                decoder_layer=decoder_layer,
+                num_layers=num_decoder_layers,
+                norm=decoder_norm,
+            )
+
+            self.decoder_output_layer = torch.nn.Linear(d_model, self.decoder_num_class)
+
+            self.decoder_criterion = LabelSmoothingLoss()
+        else:
+            self.decoder_criterion = None
+
+    def forward(
+        self, x: torch.Tensor, supervision: Optional[Supervisions] = None
+    ) -> Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (N, T, C).
+          supervision:
+            Supervision in lhotse format.
+            See https://github.com/lhotse-speech/lhotse/blob/master/lhotse/dataset/speech_recognition.py#L32  # noqa
+            (CAUTION: It contains length information, i.e., start and number of
+             frames, before subsampling)
+
+        Returns:
+          Return a tuple containing 3 tensors:
+            - CTC output for ctc decoding. Its shape is (N, T, C)
+            - Encoder output with shape (T, N, C). It can be used as key and
+              value for the decoder.
+            - Encoder output padding mask. It can be used as
+              memory_key_padding_mask for the decoder. Its shape is (N, T).
+              It is None if `supervision` is None.
+        """
+        if isinstance(self.use_feat_batchnorm, bool) and self.use_feat_batchnorm:
+            x = x.permute(0, 2, 1)  # (N, T, C) -> (N, C, T)
+            x = self.feat_batchnorm(x)
+            x = x.permute(0, 2, 1)  # (N, C, T) -> (N, T, C)
+        if isinstance(self.use_feat_batchnorm, float):
+            x *= self.use_feat_batchnorm
+        encoder_memory, memory_key_padding_mask = self.run_encoder(x, supervision)
+        x = self.ctc_output(encoder_memory)
+        return x, encoder_memory, memory_key_padding_mask
+
+    def run_encoder(
+        self, x: torch.Tensor, supervisions: Optional[Supervisions] = None
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        """Run the transformer encoder.
+
+        Args:
+          x:
+            The model input. Its shape is (N, T, C).
+          supervisions:
+            Supervision in lhotse format.
+            See https://github.com/lhotse-speech/lhotse/blob/master/lhotse/dataset/speech_recognition.py#L32  # noqa
+            CAUTION: It contains length information, i.e., start and number of
+            frames, before subsampling
+            It is read directly from the batch, without any sorting. It is used
+            to compute the encoder padding mask, which is used as memory key
+            padding mask for the decoder.
+        Returns:
+          Return a tuple with two tensors:
+            - The encoder output, with shape (T, N, C)
+            - encoder padding mask, with shape (N, T).
+              The mask is None if `supervisions` is None.
+              It is used as memory key padding mask in the decoder.
+        """
+        x = self.encoder_embed(x)
+        x = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+        mask = encoder_padding_mask(x.size(0), supervisions)
+        mask = mask.to(x.device) if mask is not None else None
+        x = self.encoder(x, src_key_padding_mask=mask)  # (T, N, C)
+
+        return x, mask
+
+    def ctc_output(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+          x:
+            The output tensor from the transformer encoder.
+            Its shape is (T, N, C)
+
+        Returns:
+          Return a tensor that can be used for CTC decoding.
+          Its shape is (N, T, C)
+        """
+        x = self.encoder_output_layer(x)
+        x = x.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+        x = nn.functional.log_softmax(x, dim=-1)  # (N, T, C)
+        return x
+
+    @torch.jit.export
+    def decoder_forward(
+        self,
+        memory: torch.Tensor,
+        memory_key_padding_mask: torch.Tensor,
+        token_ids: List[List[int]],
+        sos_id: int,
+        eos_id: int,
+    ) -> torch.Tensor:
+        """
+        Args:
+          memory:
+            It's the output of the encoder with shape (T, N, C)
+          memory_key_padding_mask:
+            The padding mask from the encoder.
+          token_ids:
+            A list-of-list IDs. Each sublist contains IDs for an utterance.
+            The IDs can be either phone IDs or word piece IDs.
+          sos_id:
+            sos token id
+          eos_id:
+            eos token id
+
+        Returns:
+            A scalar, the **sum** of label smoothing loss over utterances
+            in the batch without any normalization.
+        """
+        ys_in = add_sos(token_ids, sos_id=sos_id)
+        ys_in = [torch.tensor(y) for y in ys_in]
+        ys_in_pad = pad_sequence(ys_in, batch_first=True, padding_value=float(eos_id))
+
+        ys_out = add_eos(token_ids, eos_id=eos_id)
+        ys_out = [torch.tensor(y) for y in ys_out]
+        ys_out_pad = pad_sequence(ys_out, batch_first=True, padding_value=float(-1))
+
+        device = memory.device
+        ys_in_pad = ys_in_pad.to(device)
+        ys_out_pad = ys_out_pad.to(device)
+
+        tgt_mask = generate_square_subsequent_mask(ys_in_pad.shape[-1]).to(device)
+
+        tgt_key_padding_mask = decoder_padding_mask(ys_in_pad, ignore_id=eos_id)
+        # TODO: Use length information to create the decoder padding mask
+        # We set the first column to False since the first column in ys_in_pad
+        # contains sos_id, which is the same as eos_id in our current setting.
+        tgt_key_padding_mask[:, 0] = False
+
+        tgt = self.decoder_embed(ys_in_pad)  # (N, T) -> (N, T, C)
+        tgt = self.decoder_pos(tgt)
+        tgt = tgt.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+        pred_pad = self.decoder(
+            tgt=tgt,
+            memory=memory,
+            tgt_mask=tgt_mask,
+            tgt_key_padding_mask=tgt_key_padding_mask,
+            memory_key_padding_mask=memory_key_padding_mask,
+        )  # (T, N, C)
+        pred_pad = pred_pad.permute(1, 0, 2)  # (T, N, C) -> (N, T, C)
+        pred_pad = self.decoder_output_layer(pred_pad)  # (N, T, C)
+
+        decoder_loss = self.decoder_criterion(pred_pad, ys_out_pad)
+
+        return decoder_loss
+
+    @torch.jit.export
+    def decoder_nll(
+        self,
+        memory: torch.Tensor,
+        memory_key_padding_mask: torch.Tensor,
+        token_ids: List[torch.Tensor],
+        sos_id: int,
+        eos_id: int,
+    ) -> torch.Tensor:
+        """
+        Args:
+          memory:
+            It's the output of the encoder with shape (T, N, C)
+          memory_key_padding_mask:
+            The padding mask from the encoder.
+          token_ids:
+            A list-of-list IDs (e.g., word piece IDs).
+            Each sublist represents an utterance.
+          sos_id:
+            The token ID for SOS.
+          eos_id:
+            The token ID for EOS.
+        Returns:
+            A 2-D tensor of shape (len(token_ids), max_token_length)
+            representing the cross entropy loss (i.e., negative log-likelihood).
+        """
+        # The common part between this function and decoder_forward could be
+        # extracted as a separate function.
+        if isinstance(token_ids[0], torch.Tensor):
+            # This branch is executed by torchscript in C++.
+            # See https://github.com/k2-fsa/k2/pull/870
+            # https://github.com/k2-fsa/k2/blob/3c1c18400060415b141ccea0115fd4bf0ad6234e/k2/torch/bin/attention_rescore.cu#L286
+            token_ids = [tolist(t) for t in token_ids]
+
+        ys_in = add_sos(token_ids, sos_id=sos_id)
+        ys_in = [torch.tensor(y) for y in ys_in]
+        ys_in_pad = pad_sequence(ys_in, batch_first=True, padding_value=float(eos_id))
+
+        ys_out = add_eos(token_ids, eos_id=eos_id)
+        ys_out = [torch.tensor(y) for y in ys_out]
+        ys_out_pad = pad_sequence(ys_out, batch_first=True, padding_value=float(-1))
+
+        device = memory.device
+        ys_in_pad = ys_in_pad.to(device, dtype=torch.int64)
+        ys_out_pad = ys_out_pad.to(device, dtype=torch.int64)
+
+        tgt_mask = generate_square_subsequent_mask(ys_in_pad.shape[-1]).to(device)
+
+        tgt_key_padding_mask = decoder_padding_mask(ys_in_pad, ignore_id=eos_id)
+        # TODO: Use length information to create the decoder padding mask
+        # We set the first column to False since the first column in ys_in_pad
+        # contains sos_id, which is the same as eos_id in our current setting.
+        tgt_key_padding_mask[:, 0] = False
+
+        tgt = self.decoder_embed(ys_in_pad)  # (B, T) -> (B, T, F)
+        tgt = self.decoder_pos(tgt)
+        tgt = tgt.permute(1, 0, 2)  # (B, T, F) -> (T, B, F)
+        pred_pad = self.decoder(
+            tgt=tgt,
+            memory=memory,
+            tgt_mask=tgt_mask,
+            tgt_key_padding_mask=tgt_key_padding_mask,
+            memory_key_padding_mask=memory_key_padding_mask,
+        )  # (T, B, F)
+        pred_pad = pred_pad.permute(1, 0, 2)  # (T, B, F) -> (B, T, F)
+        pred_pad = self.decoder_output_layer(pred_pad)  # (B, T, F)
+        # nll: negative log-likelihood
+        nll = torch.nn.functional.cross_entropy(
+            pred_pad.view(-1, self.decoder_num_class),
+            ys_out_pad.view(-1),
+            ignore_index=-1,
+            reduction="none",
+        )
+
+        nll = nll.view(pred_pad.shape[0], -1)
+
+        return nll
+
+
+class TransformerEncoderLayer(nn.Module):
+    """
+    Modified from torch.nn.TransformerEncoderLayer.
+    Add support of normalize_before,
+    i.e., use layer_norm before the first block.
+
+    Args:
+      d_model:
+        the number of expected features in the input (required).
+      nhead:
+        the number of heads in the multiheadattention models (required).
+      dim_feedforward:
+        the dimension of the feedforward network model (default=2048).
+      dropout:
+        the dropout value (default=0.1).
+      activation:
+        the activation function of intermediate layer, relu or
+        gelu (default=relu).
+      normalize_before:
+        whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = TransformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> out = encoder_layer(src)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        activation: str = "relu",
+        normalize_before: bool = True,
+    ) -> None:
+        super(TransformerEncoderLayer, self).__init__()
+        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=0.0)
+        # Implementation of Feedforward model
+        self.linear1 = nn.Linear(d_model, dim_feedforward)
+        self.dropout = nn.Dropout(dropout)
+        self.linear2 = nn.Linear(dim_feedforward, d_model)
+
+        self.norm1 = nn.LayerNorm(d_model)
+        self.norm2 = nn.LayerNorm(d_model)
+        self.dropout1 = nn.Dropout(dropout)
+        self.dropout2 = nn.Dropout(dropout)
+
+        self.activation = _get_activation_fn(activation)
+
+        self.normalize_before = normalize_before
+
+    def __setstate__(self, state):
+        if "activation" not in state:
+            state["activation"] = nn.functional.relu
+        super(TransformerEncoderLayer, self).__setstate__(state)
+
+    def forward(
+        self,
+        src: torch.Tensor,
+        src_mask: Optional[torch.Tensor] = None,
+        src_key_padding_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional)
+
+        Shape:
+            src: (S, N, E).
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length,
+            N is the batch size, E is the feature number
+        """
+        residual = src
+        if self.normalize_before:
+            src = self.norm1(src)
+        src2 = self.self_attn(
+            src,
+            src,
+            src,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout1(src2)
+        if not self.normalize_before:
+            src = self.norm1(src)
+
+        residual = src
+        if self.normalize_before:
+            src = self.norm2(src)
+        src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))
+        src = residual + self.dropout2(src2)
+        if not self.normalize_before:
+            src = self.norm2(src)
+        return src
+
+
+class TransformerDecoderLayer(nn.Module):
+    """
+    Modified from torch.nn.TransformerDecoderLayer.
+    Add support of normalize_before,
+    i.e., use layer_norm before the first block.
+
+    Args:
+      d_model:
+        the number of expected features in the input (required).
+      nhead:
+        the number of heads in the multiheadattention models (required).
+      dim_feedforward:
+        the dimension of the feedforward network model (default=2048).
+      dropout:
+        the dropout value (default=0.1).
+      activation:
+        the activation function of intermediate layer, relu or
+        gelu (default=relu).
+
+    Examples::
+        >>> decoder_layer = nn.TransformerDecoderLayer(d_model=512, nhead=8)
+        >>> memory = torch.rand(10, 32, 512)
+        >>> tgt = torch.rand(20, 32, 512)
+        >>> out = decoder_layer(tgt, memory)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        activation: str = "relu",
+        normalize_before: bool = True,
+    ) -> None:
+        super(TransformerDecoderLayer, self).__init__()
+        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=0.0)
+        self.src_attn = nn.MultiheadAttention(d_model, nhead, dropout=0.0)
+        # Implementation of Feedforward model
+        self.linear1 = nn.Linear(d_model, dim_feedforward)
+        self.dropout = nn.Dropout(dropout)
+        self.linear2 = nn.Linear(dim_feedforward, d_model)
+
+        self.norm1 = nn.LayerNorm(d_model)
+        self.norm2 = nn.LayerNorm(d_model)
+        self.norm3 = nn.LayerNorm(d_model)
+        self.dropout1 = nn.Dropout(dropout)
+        self.dropout2 = nn.Dropout(dropout)
+        self.dropout3 = nn.Dropout(dropout)
+
+        self.activation = _get_activation_fn(activation)
+
+        self.normalize_before = normalize_before
+
+    def __setstate__(self, state):
+        if "activation" not in state:
+            state["activation"] = nn.functional.relu
+        super(TransformerDecoderLayer, self).__setstate__(state)
+
+    def forward(
+        self,
+        tgt: torch.Tensor,
+        memory: torch.Tensor,
+        tgt_mask: Optional[torch.Tensor] = None,
+        memory_mask: Optional[torch.Tensor] = None,
+        tgt_key_padding_mask: Optional[torch.Tensor] = None,
+        memory_key_padding_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """Pass the inputs (and mask) through the decoder layer.
+
+        Args:
+          tgt:
+            the sequence to the decoder layer (required).
+          memory:
+            the sequence from the last layer of the encoder (required).
+          tgt_mask:
+            the mask for the tgt sequence (optional).
+          memory_mask:
+            the mask for the memory sequence (optional).
+          tgt_key_padding_mask:
+            the mask for the tgt keys per batch (optional).
+          memory_key_padding_mask:
+            the mask for the memory keys per batch (optional).
+
+        Shape:
+            tgt: (T, N, E).
+            memory: (S, N, E).
+            tgt_mask: (T, T).
+            memory_mask: (T, S).
+            tgt_key_padding_mask: (N, T).
+            memory_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length,
+            N is the batch size, E is the feature number
+        """
+        residual = tgt
+        if self.normalize_before:
+            tgt = self.norm1(tgt)
+        tgt2 = self.self_attn(
+            tgt,
+            tgt,
+            tgt,
+            attn_mask=tgt_mask,
+            key_padding_mask=tgt_key_padding_mask,
+        )[0]
+        tgt = residual + self.dropout1(tgt2)
+        if not self.normalize_before:
+            tgt = self.norm1(tgt)
+
+        residual = tgt
+        if self.normalize_before:
+            tgt = self.norm2(tgt)
+        tgt2 = self.src_attn(
+            tgt,
+            memory,
+            memory,
+            attn_mask=memory_mask,
+            key_padding_mask=memory_key_padding_mask,
+        )[0]
+        tgt = residual + self.dropout2(tgt2)
+        if not self.normalize_before:
+            tgt = self.norm2(tgt)
+
+        residual = tgt
+        if self.normalize_before:
+            tgt = self.norm3(tgt)
+        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))
+        tgt = residual + self.dropout3(tgt2)
+        if not self.normalize_before:
+            tgt = self.norm3(tgt)
+        return tgt
+
+
+def _get_activation_fn(activation: str):
+    if activation == "relu":
+        return nn.functional.relu
+    elif activation == "gelu":
+        return nn.functional.gelu
+
+    raise RuntimeError("activation should be relu/gelu, not {}".format(activation))
+
+
+class PositionalEncoding(nn.Module):
+    """This class implements the positional encoding
+    proposed in the following paper:
+
+    - Attention Is All You Need: https://arxiv.org/pdf/1706.03762.pdf
+
+        PE(pos, 2i) = sin(pos / (10000^(2i/d_modle))
+        PE(pos, 2i+1) = cos(pos / (10000^(2i/d_modle))
+
+    Note::
+
+      1 / (10000^(2i/d_model)) = exp(-log(10000^(2i/d_model)))
+                               = exp(-1* 2i / d_model * log(100000))
+                               = exp(2i * -(log(10000) / d_model))
+    """
+
+    def __init__(self, d_model: int, dropout: float = 0.1) -> None:
+        """
+        Args:
+          d_model:
+            Embedding dimension.
+          dropout:
+            Dropout probability to be applied to the output of this module.
+        """
+        super().__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = nn.Dropout(p=dropout)
+        # not doing: self.pe = None because of errors thrown by torchscript
+        self.pe = torch.zeros(1, 0, self.d_model, dtype=torch.float32)
+
+    def extend_pe(self, x: torch.Tensor) -> None:
+        """Extend the time t in the positional encoding if required.
+
+        The shape of `self.pe` is (1, T1, d_model). The shape of the input x
+        is (N, T, d_model). If T > T1, then we change the shape of self.pe
+        to (N, T, d_model). Otherwise, nothing is done.
+
+        Args:
+          x:
+            It is a tensor of shape (N, T, C).
+        Returns:
+          Return None.
+        """
+        if self.pe is not None:
+            if self.pe.size(1) >= x.size(1):
+                self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        pe = torch.zeros(x.size(1), self.d_model, dtype=torch.float32)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe[:, 0::2] = torch.sin(position * div_term)
+        pe[:, 1::2] = torch.cos(position * div_term)
+        pe = pe.unsqueeze(0)
+        # Now pe is of shape (1, T, d_model), where T is x.size(1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Add positional encoding.
+
+        Args:
+          x:
+            Its shape is (N, T, C)
+
+        Returns:
+          Return a tensor of shape (N, T, C)
+        """
+        self.extend_pe(x)
+        x = x * self.xscale + self.pe[:, : x.size(1), :]
+        return self.dropout(x)
+
+
+class Noam(object):
+    """
+    Implements Noam optimizer.
+
+    Proposed in
+    "Attention Is All You Need", https://arxiv.org/pdf/1706.03762.pdf
+
+    Modified from
+    https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/optimizer.py  # noqa
+
+    Args:
+      params:
+        iterable of parameters to optimize or dicts defining parameter groups
+      model_size:
+        attention dimension of the transformer model
+      factor:
+        learning rate factor
+      warm_step:
+        warmup steps
+    """
+
+    def __init__(
+        self,
+        params,
+        model_size: int = 256,
+        factor: float = 10.0,
+        warm_step: int = 25000,
+        weight_decay=0,
+    ) -> None:
+        """Construct an Noam object."""
+        self.optimizer = torch.optim.Adam(
+            params, lr=0, betas=(0.9, 0.98), eps=1e-9, weight_decay=weight_decay
+        )
+        self._step = 0
+        self.warmup = warm_step
+        self.factor = factor
+        self.model_size = model_size
+        self._rate = 0
+
+    @property
+    def param_groups(self):
+        """Return param_groups."""
+        return self.optimizer.param_groups
+
+    def step(self):
+        """Update parameters and rate."""
+        self._step += 1
+        rate = self.rate()
+        for p in self.optimizer.param_groups:
+            p["lr"] = rate
+        self._rate = rate
+        self.optimizer.step()
+
+    def rate(self, step=None):
+        """Implement `lrate` above."""
+        if step is None:
+            step = self._step
+        return (
+            self.factor
+            * self.model_size ** (-0.5)
+            * min(step ** (-0.5), step * self.warmup ** (-1.5))
+        )
+
+    def zero_grad(self):
+        """Reset gradient."""
+        self.optimizer.zero_grad()
+
+    def state_dict(self):
+        """Return state_dict."""
+        return {
+            "_step": self._step,
+            "warmup": self.warmup,
+            "factor": self.factor,
+            "model_size": self.model_size,
+            "_rate": self._rate,
+            "optimizer": self.optimizer.state_dict(),
+        }
+
+    def load_state_dict(self, state_dict):
+        """Load state_dict."""
+        for key, value in state_dict.items():
+            if key == "optimizer":
+                self.optimizer.load_state_dict(state_dict["optimizer"])
+            else:
+                setattr(self, key, value)
+
+
+def encoder_padding_mask(
+    max_len: int, supervisions: Optional[Supervisions] = None
+) -> Optional[torch.Tensor]:
+    """Make mask tensor containing indexes of padded part.
+
+    TODO::
+      This function **assumes** that the model uses
+      a subsampling factor of 4. We should remove that
+      assumption later.
+
+    Args:
+      max_len:
+        Maximum length of input features.
+        CAUTION: It is the length after subsampling.
+      supervisions:
+        Supervision in lhotse format.
+        See https://github.com/lhotse-speech/lhotse/blob/master/lhotse/dataset/speech_recognition.py#L32  # noqa
+        (CAUTION: It contains length information, i.e., start and number of
+         frames, before subsampling)
+
+    Returns:
+        Tensor: Mask tensor of dimension (batch_size, input_length),
+        True denote the masked indices.
+    """
+    if supervisions is None:
+        return None
+
+    supervision_segments = torch.stack(
+        (
+            supervisions["sequence_idx"],
+            supervisions["start_frame"],
+            supervisions["num_frames"],
+        ),
+        1,
+    ).to(torch.int32)
+
+    lengths = [0 for _ in range(int(supervision_segments[:, 0].max().item()) + 1)]
+    for idx in range(supervision_segments.size(0)):
+        # Note: TorchScript doesn't allow to unpack tensors as tuples
+        sequence_idx = supervision_segments[idx, 0].item()
+        start_frame = supervision_segments[idx, 1].item()
+        num_frames = supervision_segments[idx, 2].item()
+        lengths[sequence_idx] = start_frame + num_frames
+
+    lengths = [((i - 1) // 2 - 1) // 2 for i in lengths]
+    bs = int(len(lengths))
+    seq_range = torch.arange(0, max_len, dtype=torch.int64)
+    seq_range_expand = seq_range.unsqueeze(0).expand(bs, max_len)
+    # Note: TorchScript doesn't implement Tensor.new()
+    seq_length_expand = torch.tensor(
+        lengths, device=seq_range_expand.device, dtype=seq_range_expand.dtype
+    ).unsqueeze(-1)
+    mask = seq_range_expand >= seq_length_expand
+
+    return mask
+
+
+def decoder_padding_mask(ys_pad: torch.Tensor, ignore_id: int = -1) -> torch.Tensor:
+    """Generate a length mask for input.
+
+    The masked position are filled with True,
+    Unmasked positions are filled with False.
+
+    Args:
+      ys_pad:
+        padded tensor of dimension (batch_size, input_length).
+      ignore_id:
+        the ignored number (the padding number) in ys_pad
+
+    Returns:
+      Tensor:
+        a bool tensor of the same shape as the input tensor.
+    """
+    ys_mask = ys_pad == ignore_id
+    return ys_mask
+
+
+def generate_square_subsequent_mask(sz: int) -> torch.Tensor:
+    """Generate a square mask for the sequence. The masked positions are
+    filled with float('-inf'). Unmasked positions are filled with float(0.0).
+    The mask can be used for masked self-attention.
+
+    For instance, if sz is 3, it returns::
+
+        tensor([[0., -inf, -inf],
+                [0., 0., -inf],
+                [0., 0., 0]])
+
+    Args:
+      sz: mask size
+
+    Returns:
+      A square mask of dimension (sz, sz)
+    """
+    mask = (torch.triu(torch.ones(sz, sz)) == 1).transpose(0, 1)
+    mask = (
+        mask.float()
+        .masked_fill(mask == 0, float("-inf"))
+        .masked_fill(mask == 1, float(0.0))
+    )
+    return mask
+
+
+def add_sos(token_ids: List[List[int]], sos_id: int) -> List[List[int]]:
+    """Prepend sos_id to each utterance.
+
+    Args:
+      token_ids:
+        A list-of-list of token IDs. Each sublist contains
+        token IDs (e.g., word piece IDs) of an utterance.
+      sos_id:
+        The ID of the SOS token.
+
+    Return:
+      Return a new list-of-list, where each sublist starts
+      with SOS ID.
+    """
+    return [[sos_id] + utt for utt in token_ids]
+
+
+def add_eos(token_ids: List[List[int]], eos_id: int) -> List[List[int]]:
+    """Append eos_id to each utterance.
+
+    Args:
+      token_ids:
+        A list-of-list of token IDs. Each sublist contains
+        token IDs (e.g., word piece IDs) of an utterance.
+      eos_id:
+        The ID of the EOS token.
+
+    Return:
+      Return a new list-of-list, where each sublist ends
+      with EOS ID.
+    """
+    return [utt + [eos_id] for utt in token_ids]
+
+
+def tolist(t: torch.Tensor) -> List[int]:
+    """Used by jit"""
+    return torch.jit.annotate(List[int], t.tolist())

egs/swbd/ASR/local/compile_hlg.py

@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This script takes as input lang_dir and generates HLG from
+
+    - H, the ctc topology, built from tokens contained in lang_dir/lexicon.txt
+    - L, the lexicon, built from lang_dir/L_disambig.pt
+
+        Caution: We use a lexicon that contains disambiguation symbols
+
+    - G, the LM, built from data/lm/G_n_gram.fst.txt
+
+The generated HLG is saved in $lang_dir/HLG.pt
+"""
+import argparse
+import logging
+from pathlib import Path
+
+import k2
+import torch
+
+from icefall.lexicon import Lexicon
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lm",
+        type=str,
+        default="G_3_gram",
+        help="""Stem name for LM used in HLG compiling.
+        """,
+    )
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        """,
+    )
+
+    return parser.parse_args()
+
+
+def compile_HLG(lang_dir: str, lm: str = "G_3_gram") -> k2.Fsa:
+    """
+    Args:
+      lang_dir:
+        The language directory, e.g., data/lang_phone or data/lang_bpe_5000.
+      lm:
+        The language stem base name.
+
+    Return:
+      An FSA representing HLG.
+    """
+    lexicon = Lexicon(lang_dir)
+    max_token_id = max(lexicon.tokens)
+    logging.info(f"Building ctc_topo. max_token_id: {max_token_id}")
+    H = k2.ctc_topo(max_token_id)
+    L = k2.Fsa.from_dict(torch.load(f"{lang_dir}/L_disambig.pt"))
+
+    if Path(f"data/lm/{lm}.pt").is_file():
+        logging.info(f"Loading pre-compiled {lm}")
+        d = torch.load(f"data/lm/{lm}.pt")
+        G = k2.Fsa.from_dict(d)
+    else:
+        logging.info(f"Loading {lm}.fst.txt")
+        with open(f"data/lm/{lm}.fst.txt") as f:
+            G = k2.Fsa.from_openfst(f.read(), acceptor=False)
+            torch.save(G.as_dict(), f"data/lm/{lm}.pt")
+
+    first_token_disambig_id = lexicon.token_table["#0"]
+    first_word_disambig_id = lexicon.word_table["#0"]
+
+    L = k2.arc_sort(L)
+    G = k2.arc_sort(G)
+
+    logging.info("Intersecting L and G")
+    LG = k2.compose(L, G)
+    logging.info(f"LG shape: {LG.shape}")
+
+    logging.info("Connecting LG")
+    LG = k2.connect(LG)
+    logging.info(f"LG shape after k2.connect: {LG.shape}")
+
+    logging.info(type(LG.aux_labels))
+    logging.info("Determinizing LG")
+
+    LG = k2.determinize(LG)
+    logging.info(type(LG.aux_labels))
+
+    logging.info("Connecting LG after k2.determinize")
+    LG = k2.connect(LG)
+
+    logging.info("Removing disambiguation symbols on LG")
+
+    # LG.labels[LG.labels >= first_token_disambig_id] = 0
+    # see https://github.com/k2-fsa/k2/pull/1140
+    labels = LG.labels
+    labels[labels >= first_token_disambig_id] = 0
+    LG.labels = labels
+
+    assert isinstance(LG.aux_labels, k2.RaggedTensor)
+    LG.aux_labels.values[LG.aux_labels.values >= first_word_disambig_id] = 0
+
+    LG = k2.remove_epsilon(LG)
+    logging.info(f"LG shape after k2.remove_epsilon: {LG.shape}")
+
+    LG = k2.connect(LG)
+    LG.aux_labels = LG.aux_labels.remove_values_eq(0)
+
+    logging.info("Arc sorting LG")
+    LG = k2.arc_sort(LG)
+
+    logging.info("Composing H and LG")
+    # CAUTION: The name of the inner_labels is fixed
+    # to `tokens`. If you want to change it, please
+    # also change other places in icefall that are using
+    # it.
+    HLG = k2.compose(H, LG, inner_labels="tokens")
+
+    logging.info("Connecting LG")
+    HLG = k2.connect(HLG)
+
+    logging.info("Arc sorting LG")
+    HLG = k2.arc_sort(HLG)
+    logging.info(f"HLG.shape: {HLG.shape}")
+
+    return HLG
+
+
+def main():
+    args = get_args()
+    lang_dir = Path(args.lang_dir)
+
+    if (lang_dir / "HLG.pt").is_file():
+        logging.info(f"{lang_dir}/HLG.pt already exists - skipping")
+        return
+
+    logging.info(f"Processing {lang_dir}")
+
+    HLG = compile_HLG(lang_dir, args.lm)
+    logging.info(f"Saving HLG.pt to {lang_dir}")
+    torch.save(HLG.as_dict(), f"{lang_dir}/HLG.pt")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/swbd/ASR/local/compile_lg.py

@@ -0,0 +1,147 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang, Wei Kang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This script takes as input lang_dir and generates LG from
+
+    - L, the lexicon, built from lang_dir/L_disambig.pt
+
+        Caution: We use a lexicon that contains disambiguation symbols
+
+    - G, the LM, built from data/lm/G_3_gram.fst.txt
+
+The generated LG is saved in $lang_dir/LG.pt
+"""
+import argparse
+import logging
+from pathlib import Path
+
+import k2
+import torch
+
+from icefall.lexicon import Lexicon
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        """,
+    )
+    parser.add_argument(
+        "--lm",
+        type=str,
+        default="G_3_gram",
+        help="""Stem name for LM used in HLG compiling.
+        """,
+    )
+
+    return parser.parse_args()
+
+
+def compile_LG(lang_dir: str, lm: str = "G_3_gram") -> k2.Fsa:
+    """
+    Args:
+      lang_dir:
+        The language directory, e.g., data/lang_phone or data/lang_bpe_5000.
+
+    Return:
+      An FSA representing LG.
+    """
+    lexicon = Lexicon(lang_dir)
+    L = k2.Fsa.from_dict(torch.load(f"{lang_dir}/L_disambig.pt"))
+
+    if Path(f"data/lm/{lm}.pt").is_file():
+        logging.info(f"Loading pre-compiled {lm}")
+        d = torch.load(f"data/lm/{lm}.pt")
+        G = k2.Fsa.from_dict(d)
+    else:
+        logging.info(f"Loading {lm}.fst.txt")
+        with open(f"data/lm/{lm}.fst.txt") as f:
+            G = k2.Fsa.from_openfst(f.read(), acceptor=False)
+            torch.save(G.as_dict(), f"data/lm/{lm}.pt")
+
+    first_token_disambig_id = lexicon.token_table["#0"]
+    first_word_disambig_id = lexicon.word_table["#0"]
+
+    L = k2.arc_sort(L)
+    G = k2.arc_sort(G)
+
+    logging.info("Intersecting L and G")
+    LG = k2.compose(L, G)
+    logging.info(f"LG shape: {LG.shape}")
+
+    logging.info("Connecting LG")
+    LG = k2.connect(LG)
+    logging.info(f"LG shape after k2.connect: {LG.shape}")
+
+    logging.info(type(LG.aux_labels))
+    logging.info("Determinizing LG")
+
+    LG = k2.determinize(LG, k2.DeterminizeWeightPushingType.kLogWeightPushing)
+    logging.info(type(LG.aux_labels))
+
+    logging.info("Connecting LG after k2.determinize")
+    LG = k2.connect(LG)
+
+    logging.info("Removing disambiguation symbols on LG")
+
+    # LG.labels[LG.labels >= first_token_disambig_id] = 0
+    # see https://github.com/k2-fsa/k2/pull/1140
+    labels = LG.labels
+    labels[labels >= first_token_disambig_id] = 0
+    LG.labels = labels
+
+    assert isinstance(LG.aux_labels, k2.RaggedTensor)
+    LG.aux_labels.values[LG.aux_labels.values >= first_word_disambig_id] = 0
+
+    LG = k2.remove_epsilon(LG)
+    logging.info(f"LG shape after k2.remove_epsilon: {LG.shape}")
+
+    LG = k2.connect(LG)
+    LG.aux_labels = LG.aux_labels.remove_values_eq(0)
+
+    logging.info("Arc sorting LG")
+    LG = k2.arc_sort(LG)
+
+    return LG
+
+
+def main():
+    args = get_args()
+    lang_dir = Path(args.lang_dir)
+
+    if (lang_dir / "LG.pt").is_file():
+        logging.info(f"{lang_dir}/LG.pt already exists - skipping")
+        return
+
+    logging.info(f"Processing {lang_dir}")
+
+    LG = compile_LG(lang_dir, args.lm)
+    logging.info(f"Saving LG.pt to {lang_dir}")
+    torch.save(LG.as_dict(), f"{lang_dir}/LG.pt")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/swbd/ASR/local/compute_fbank_swbd.py

@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# Modified     2023  The Chinese University of Hong Kong (author: Zengrui Jin)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This file computes fbank features of the SwitchBoard dataset.
+It looks for manifests in the directory data/manifests.
+
+The generated fbank features are saved in data/fbank.
+"""
+
+import argparse
+import logging
+import os
+from pathlib import Path
+from typing import Optional
+
+import sentencepiece as spm
+import torch
+from filter_cuts import filter_cuts
+from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
+from lhotse.recipes.utils import read_manifests_if_cached
+
+from icefall.utils import get_executor, str2bool
+
+# Torch's multithreaded behavior needs to be disabled or
+# it wastes a lot of CPU and slow things down.
+# Do this outside of main() in case it needs to take effect
+# even when we are not invoking the main (e.g. when spawning subprocesses).
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        help="""Path to the bpe.model. If not None, we will remove short and
+        long utterances before extracting features""",
+    )
+
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        help="""Dataset parts to compute fbank. If None, we will use all""",
+    )
+
+    parser.add_argument(
+        "--perturb-speed",
+        type=str2bool,
+        default=True,
+        help="""Perturb speed with factor 0.9 and 1.1 on train subset.""",
+    )
+
+    return parser.parse_args()
+
+
+def compute_fbank_switchboard(
+    bpe_model: Optional[str] = None,
+    dataset: Optional[str] = None,
+    perturb_speed: Optional[bool] = True,
+):
+    src_dir = Path("data/manifests")
+    output_dir = Path("data/fbank")
+    num_jobs = min(15, os.cpu_count())
+    num_mel_bins = 80
+
+    if bpe_model:
+        logging.info(f"Loading {bpe_model}")
+        sp = spm.SentencePieceProcessor()
+        sp.load(bpe_model)
+
+    if dataset is None:
+        dataset_parts = ("all", "eval2000", "rt03")
+    else:
+        dataset_parts = dataset.split(" ", -1)
+
+    prefix = "swbd"
+    suffix = "jsonl.gz"
+    manifests = read_manifests_if_cached(
+        dataset_parts=dataset_parts,
+        output_dir=src_dir,
+        prefix=prefix,
+        suffix=suffix,
+    )
+    assert manifests is not None
+
+    assert len(manifests) == len(dataset_parts), (
+        len(manifests),
+        len(dataset_parts),
+        list(manifests.keys()),
+        dataset_parts,
+    )
+
+    extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
+
+    with get_executor() as ex:  # Initialize the executor only once.
+        for partition, m in manifests.items():
+            cuts_filename = f"{prefix}_cuts_{partition}.{suffix}"
+            if (output_dir / cuts_filename).is_file():
+                logging.info(f"{partition} already exists - skipping.")
+                continue
+            logging.info(f"Processing {partition}")
+            cut_set = CutSet.from_manifests(
+                recordings=m["recordings"].resample(16000),
+                supervisions=m["supervisions"],
+            )
+
+            if "train" in partition:
+                if bpe_model:
+                    cut_set = filter_cuts(cut_set, sp)
+                if perturb_speed:
+                    logging.info(f"Doing speed perturb")
+                    cut_set = (
+                        cut_set
+                        + cut_set.perturb_speed(0.9)
+                        + cut_set.perturb_speed(1.1)
+                    ).resample(16000)
+            cut_set = cut_set.compute_and_store_features(
+                extractor=extractor,
+                storage_path=f"{output_dir}/{prefix}_feats_{partition}",
+                # when an executor is specified, make more partitions
+                num_jobs=num_jobs if ex is None else 80,
+                executor=ex,
+                storage_type=LilcomChunkyWriter,
+            )
+            cut_set.to_file(output_dir / cuts_filename)
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    args = get_args()
+    logging.info(vars(args))
+    compute_fbank_switchboard(
+        bpe_model=args.bpe_model,
+        dataset=args.dataset,
+        perturb_speed=args.perturb_speed,
+    )

egs/swbd/ASR/local/convert_transcript_words_to_tokens.py

@@ -0,0 +1,103 @@
+#!/usr/bin/env python3
+
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+"""
+Convert a transcript file containing words to a corpus file containing tokens
+for LM training with the help of a lexicon.
+
+If the lexicon contains phones, the resulting LM will be a phone LM; If the
+lexicon contains word pieces, the resulting LM will be a word piece LM.
+
+If a word has multiple pronunciations, the one that appears first in the lexicon
+is kept; others are removed.
+
+If the input transcript is:
+
+    hello zoo world hello
+    world zoo
+    foo zoo world hellO
+
+and if the lexicon is
+
+    <UNK> SPN
+    hello h e l l o 2
+    hello h e l l o
+    world w o r l d
+    zoo z o o
+
+Then the output is
+
+    h e l l o 2 z o o w o r l d h e l l o 2
+    w o r l d z o o
+    SPN z o o w o r l d SPN
+"""
+
+import argparse
+from pathlib import Path
+from typing import Dict, List
+
+from generate_unique_lexicon import filter_multiple_pronunications
+
+from icefall.lexicon import read_lexicon
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--transcript",
+        type=str,
+        help="The input transcript file."
+        "We assume that the transcript file consists of "
+        "lines. Each line consists of space separated words.",
+    )
+    parser.add_argument("--lexicon", type=str, help="The input lexicon file.")
+    parser.add_argument("--oov", type=str, default="<UNK>", help="The OOV word.")
+
+    return parser.parse_args()
+
+
+def process_line(lexicon: Dict[str, List[str]], line: str, oov_token: str) -> None:
+    """
+    Args:
+      lexicon:
+        A dict containing pronunciations. Its keys are words and values
+        are pronunciations (i.e., tokens).
+      line:
+        A line of transcript consisting of space(s) separated words.
+      oov_token:
+        The pronunciation of the oov word if a word in `line` is not present
+        in the lexicon.
+    Returns:
+      Return None.
+    """
+    s = ""
+    words = line.strip().split()
+    for i, w in enumerate(words):
+        tokens = lexicon.get(w, oov_token)
+        s += " ".join(tokens)
+        s += " "
+    print(s.strip())
+
+
+def main():
+    args = get_args()
+    assert Path(args.lexicon).is_file()
+    assert Path(args.transcript).is_file()
+    assert len(args.oov) > 0
+
+    # Only the first pronunciation of a word is kept
+    lexicon = filter_multiple_pronunications(read_lexicon(args.lexicon))
+
+    lexicon = dict(lexicon)
+
+    assert args.oov in lexicon
+
+    oov_token = lexicon[args.oov]
+
+    with open(args.transcript) as f:
+        for line in f:
+            process_line(lexicon=lexicon, line=line, oov_token=oov_token)
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/local/dict.patch

@@ -0,0 +1,380 @@
+1d0
+<  file: $SWB/data/dictionary/sw-ms98-dict.text
+8645a8646
+> uh-hum ah m hh ah m
+9006c9007
+< April ey p r ih l
+---
+> April ey p r ax l
+9144d9144
+< B ay zh aa n iy z
+9261c9261
+< Battle b ae t el
+---
+> Battle b ae t ax l
+10014a10015
+> Chevy sh eh v iy
+10211a10213
+> Colorado k ao l ax r aa d ow
+10212a10215
+> Colorado' k ao l ax r aa d ow z
+10370c10373
+< Creek k r ih k
+---
+> Creek k r iy k
+10889a10893
+> Eleven ax l eh v ih n
+10951c10955
+< Erie ih r iy
+---
+> Erie iy r iy
+11183c11187
+< Forever f ax r eh v er
+---
+> Forever f er eh v er
+11231a11236
+> Friday f r ay d iy
+11744a11750
+> History hh ih s t r iy
+12004a12011,12012
+> Israel ih z r ih l
+> Israel's ih z r ih l z
+12573a12582
+> Lincoln l ih ng k ih n
+12574a12584
+> Lincolns l ih ng k ih n z
+13268c13278
+< NAACP eh ey ey s iy p iy
+---
+> NAACP eh n ey ey s iy p iy
+13286c13296
+< NIT eh ay t iy
+---
+> NIT eh n ay t iy
+13292c13302
+< NTSC eh t iy eh s s iy
+---
+> NTSC eh n t iy eh s s iy
+14058a14069
+> Quarter k ow r t er
+14059a14071
+> Quarterback k ow r t er b ae k
+14060a14073
+> Quarters k ow r t er z
+14569a14583
+> Science s ay n s
+15087a15102
+> Sunday s ah n d iy
+15088a15104
+> Sunday's s ah n d iy z
+15089a15106
+> Sundays s ah n d iy z
+15290,15291c15307,15308
+< Texan t eh k sh ih n
+< Texan's t eh k sh ih n s
+---
+> Texan t eh k s ih n
+> Texan's t eh k s ih n s
+15335a15353
+> Thousands th aw z ih n z
+15739c15757
+< Waco w ae k ow
+---
+> Waco w ey k ow
+15841a15860
+> Weekends w iy k eh n z
+16782a16802
+> acceptable eh k s eh p ax b ax l
+16833a16854
+> accounting ax k aw n ih ng
+16948a16970
+> address ax d r eh s
+17281a17304
+> already aa r d iy
+17315a17339
+> am m
+17709a17734
+> asked ae s t
+17847a17873
+> attorney ih t er n iy
+17919a17946
+> autopilot ao t ow p ay l ih t
+17960a17988
+> awfully ao f l iy
+18221a18250
+> basketball b ae s k ax b ao l
+18222a18252
+> basketball's b ae s k ax b ao l z
+18302a18333
+> become b ah k ah m
+18303a18335
+> becomes b iy k ah m z
+18344a18377
+> began b ax g en n
+18817c18850
+< bottle b aa t el
+---
+> bottle b aa t ax l
+19332,19333c19365,19367
+< camera's k ae m ax r ax z
+< cameras k ae m ax r ax z
+---
+> camera k ae m r ax
+> camera's k ae m r ax z
+> cameras k ae m r ax z
+19411a19446
+> capital k ae p ax l
+19505a19541
+> carrying k ae r ih ng
+20316a20353,20354
+> combination k aa m ih n ey sh ih n
+> combinations k aa m ih n ey sh ih n z
+20831a20870
+> contracts k aa n t r ae k s
+21010a21050
+> costs k ao s
+21062a21103
+> county k aw n iy
+21371a21413
+> cultural k ao l ch ax r ax l
+21372a21415
+> culturally k ao l ch ax r ax l iy
+21373a21417
+> culture k ao l ch er
+21375a21420
+> cultures k ao l ch er z
+21543a21589
+> data d ey t ax
+22097a22144
+> differently d ih f ax r ih n t l iy
+22972a23020
+> effects ax f eh k t s
+23016a23065
+> election ax l eh k sh ih n
+23018a23068
+> elections ax l eh k sh ih n z
+23052a23103
+> eleven ax l eh v ih n
+23242a23294
+> enjoyable ae n jh oy ax b ax l
+23248a23301
+> enjoys ae n jh oy z
+23293a23347
+> entire ih n t ay r
+23295a23350,23351
+> entirely ih n t ay r l iy
+> entirety ih n t ay r t iy
+23745a23802
+> extra eh k s t er
+23818a23876
+> facts f ae k s
+24508c24566
+< forever f ax r eh v er
+---
+> forever f er eh v er
+24514c24572
+< forget f ow r g eh t
+---
+> forget f er r g eh t
+24521a24580
+> forgot f er r g aa t
+24522a24582
+> forgotten f er r g aa t ax n
+24563a24624
+> forward f ow er d
+24680a24742
+> frightening f r ay t n ih ng
+24742a24805
+> full-time f ax l t ay m
+24862a24926
+> garage g r aa jh
+25218a25283
+> grandmother g r ae m ah dh er
+25790a25856
+> heavily hh eh v ax l iy
+25949a26016
+> history hh ih s t r iy
+26038a26106
+> honestly aa n ax s t l iy
+26039a26108
+> honesty aa n ax s t iy
+26099a26169
+> horror hh ow r
+26155a26226
+> houses hh aw z ih z
+26184c26255
+< huh-uh hh ah hh ah
+---
+> huh-uh ah hh ah
+26189c26260
+< hum-um hh m hh m
+---
+> hum-um ah m hh ah m
+26236a26308
+> hunting hh ah n ih ng
+26307a26380,26381
+> ideal ay d iy l
+> idealist ay d iy l ih s t
+26369a26444
+> imagine m ae jh ih n
+26628a26704
+> individuals ih n d ih v ih jh ax l z
+26968a27045
+> interest ih n t r ih s t
+27184a27262
+> it'd ih d
+27702a27781
+> lead l iy d
+28378a28458
+> mandatory m ae n d ih t ow r iy
+28885a28966
+> minute m ih n ih t
+29167a29249
+> mountains m aw t n z
+29317a29400
+> mysteries m ih s t r iy z
+29318a29402
+> mystery m ih s t r iy
+29470a29555
+> nervous n er v ih s
+29578,29580c29663,29665
+< nobody n ow b aa d iy
+< nobody'll n ow b aa d iy l
+< nobody's n ow b aa d iy z
+---
+> nobody n ow b ah d iy
+> nobody'll n ow b ah d iy l
+> nobody's n ow b ah d iy z
+29712a29798
+> nuclear n uw k l iy r
+29938a30025
+> onto aa n t ax
+30051a30139
+> originally ax r ih jh ax l iy
+30507a30596
+> particularly p er t ih k y ax l iy
+30755a30845
+> perfectly p er f ih k l iy
+30820a30911
+> personally p er s n ax l iy
+30915a31007
+> physically f ih z ih k l iy
+30986a31079
+> pilot p ay l ih t
+30987a31081
+> pilot's p ay l ih t s
+31227a31322
+> police p l iy s
+31513a31609
+> prefer p er f er
+31553a31650
+> prepare p r ax p ey r
+31578a31676
+> prescription p er s k r ih p sh ih n
+31579a31678
+> prescriptions p er s k r ih p sh ih n z
+31770a31870
+> products p r aa d ax k s
+31821a31922
+> projects p r aa jh eh k s
+31908a32010
+> protect p er t eh k t
+31909a32012
+> protected p er t eh k t ih d
+31911a32015
+> protection p er t eh k sh ih n
+31914a32019
+> protection p er t eh k t ih v
+32149a32255
+> quarter k ow r t er
+32414a32521
+> read r iy d
+32785a32893
+> rehabilitation r iy ax b ih l ih t ey sh ih n
+33150a33259
+> resource r ih s ow r s
+33151a33261
+> resources r iy s ow r s ih z
+33539c33649
+< roots r uh t s
+---
+> roots r uw t s
+33929a34040
+> science s ay n s
+34315a34427
+> seventy s eh v ih n iy
+34319,34320c34431,34432
+< severe s ax v iy r
+< severely s ax v iy r l iy
+---
+> severe s ih v iy r
+> severely s ih v iy r l iy
+35060a35173
+> software s ao f w ey r
+35083a35197
+> solid s ao l ih d
+35084a35199
+> solidly s ao l ih d l iy
+35750a35866
+> stood s t ih d
+35854a35971
+> strictly s t r ih k l iy
+35889c36006
+< stronger s t r ao ng er
+---
+> stronger s t r ao ng g er
+36192a36310,36311
+> supposed s p ow z
+> supposed s p ow s
+36510a36630
+> tastes t ey s
+36856a36977
+> thoroughly th er r l iy
+36866a36988
+> thousands th aw z ih n z
+37081c37203
+< toots t uh t s
+---
+> toots t uw t s
+37157a37280
+> toward t w ow r d
+37158a37282
+> towards t w ow r d z
+37564a37689
+> twenties t w eh n iy z
+37565a37691
+> twentieth t w eh n iy ih th
+37637a37764
+> unacceptable ah n ae k s eh p ax b ax l
+37728a37856
+> understand ah n d er s t ae n
+37860a37989
+> unless ih n l eh s
+38040a38170
+> use y uw z
+38049a38180
+> uses y uw z ih z
+38125a38257
+> various v ah r iy ih s
+38202a38335
+> versus v er s ih z
+38381c38514
+< wacko w ae k ow
+---
+> wacko w ey k ow
+38455c38588
+< wanna w aa n ax
+---
+> wanna w ah n ax
+38675c38808
+< whatnot w ah t n aa t
+---
+> whatnot w aa t n aa t
+38676a38810
+> whatsoever w aa t s ow eh v er
+38890c39024
+< wok w aa k
+---
+> wok w ao k
+38910a39045
+> wondering w ah n d r ih ng

egs/swbd/ASR/local/display_manifest_statistics.py

@@ -0,0 +1,211 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+This file displays duration statistics of utterances in a manifest.
+You can use the displayed value to choose minimum/maximum duration
+to remove short and long utterances during the training.
+
+See the function `remove_short_and_long_utt()` in transducer/train.py
+for usage.
+"""
+
+
+from lhotse import load_manifest_lazy
+
+
+def main():
+    #  path = "./data/fbank/swbd_cuts_rt03.jsonl.gz"
+    #  path = "./data/fbank/swbd_cuts_eval2000.jsonl.gz"
+    path = "./data/fbank/swbd_cuts_all.jsonl.gz"
+
+    cuts = load_manifest_lazy(path)
+    cuts.describe()
+
+
+if __name__ == "__main__":
+    main()
+
+"""
+## train-clean-100
+Cuts count: 85617
+Total duration (hours): 303.8
+Speech duration (hours): 303.8 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.8
+std     3.8
+min     1.3
+0.1%    1.9
+0.5%    2.2
+1%      2.5
+5%      4.2
+10%     6.4
+25%     11.4
+50%     13.8
+75%     15.3
+90%     16.7
+95%     17.3
+99%     18.1
+99.5%   18.4
+99.9%   18.8
+max     27.2
+
+## train-clean-360
+Cuts count: 312042
+Total duration (hours): 1098.2
+Speech duration (hours): 1098.2 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.7
+std     3.8
+min     1.0
+0.1%    1.8
+0.5%    2.2
+1%      2.5
+5%      4.2
+10%     6.2
+25%     11.2
+50%     13.7
+75%     15.3
+90%     16.6
+95%     17.3
+99%     18.1
+99.5%   18.4
+99.9%   18.8
+max     33.0
+
+## train-other 500
+Cuts count: 446064
+Total duration (hours): 1500.6
+Speech duration (hours): 1500.6 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.1
+std     4.2
+min     0.8
+0.1%    1.7
+0.5%    2.1
+1%      2.3
+5%      3.5
+10%     5.0
+25%     9.8
+50%     13.4
+75%     15.1
+90%     16.5
+95%     17.2
+99%     18.1
+99.5%   18.4
+99.9%   18.9
+max     31.0
+
+## dev-clean
+Cuts count: 2703
+Total duration (hours): 5.4
+Speech duration (hours): 5.4 (100.0%)
+***
+Duration statistics (seconds):
+mean    7.2
+std     4.7
+min     1.4
+0.1%    1.6
+0.5%    1.8
+1%      1.9
+5%      2.4
+10%     2.7
+25%     3.8
+50%     5.9
+75%     9.3
+90%     13.3
+95%     16.4
+99%     23.8
+99.5%   28.5
+99.9%   32.3
+max     32.6
+
+## dev-other
+Cuts count: 2864
+Total duration (hours): 5.1
+Speech duration (hours): 5.1 (100.0%)
+***
+Duration statistics (seconds):
+mean    6.4
+std     4.3
+min     1.1
+0.1%    1.3
+0.5%    1.7
+1%      1.8
+5%      2.2
+10%     2.6
+25%     3.5
+50%     5.3
+75%     7.9
+90%     12.0
+95%     15.0
+99%     22.2
+99.5%   27.1
+99.9%   32.4
+max     35.2
+
+## test-clean
+Cuts count: 2620
+Total duration (hours): 5.4
+Speech duration (hours): 5.4 (100.0%)
+***
+Duration statistics (seconds):
+mean    7.4
+std     5.2
+min     1.3
+0.1%    1.6
+0.5%    1.8
+1%      2.0
+5%      2.3
+10%     2.7
+25%     3.7
+50%     5.8
+75%     9.6
+90%     14.6
+95%     17.8
+99%     25.5
+99.5%   28.4
+99.9%   32.8
+max     35.0
+
+## test-other
+Cuts count: 2939
+Total duration (hours): 5.3
+Speech duration (hours): 5.3 (100.0%)
+***
+Duration statistics (seconds):
+mean    6.5
+std     4.4
+min     1.2
+0.1%    1.5
+0.5%    1.8
+1%      1.9
+5%      2.3
+10%     2.6
+25%     3.4
+50%     5.2
+75%     8.2
+90%     12.6
+95%     15.8
+99%     21.4
+99.5%   23.8
+99.9%   33.5
+max     34.5
+"""

egs/swbd/ASR/local/eval2000_data_prep.sh

@@ -0,0 +1,120 @@
+#!/bin/bash
+
+# Hub-5 Eval 2000 data preparation
+# Author:  Arnab Ghoshal (Jan 2013)
+
+# To be run from one directory above this script.
+
+# The input is two directory names (possibly the same) containing the
+# 2000 Hub5 english evaluation test set and transcripts, which are
+# respectively: LDC2002S09  LDC2002T43
+# e.g. see
+# http://www.ldc.upenn.edu/Catalog/catalogEntry.jsp?catalogId=LDC2002S09
+# http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2002T43
+#
+# Example usage:
+# local/eval2000_data_prep_edin.sh /exports/work/inf_hcrc_cstr_general/corpora/hub5/2000 /exports/work/inf_hcrc_cstr_general/corpora/hub5/2000/transcr
+# The first directory ($sdir) contains the speech data, and the directory
+# $sdir/english/ must exist.
+# The second directory ($tdir) contains the transcripts, and the directory
+# $tdir/reference must exist; in particular we need the file
+# $tdir/reference/hub5e00.english.000405.stm
+
+if [ $# -ne 2 ]; then
+    echo "Usage: "`basename $0`" <speech-dir> <transcription-dir>"
+    echo "See comments in the script for more details"
+    exit 1
+fi
+
+sdir=$1
+tdir=$2
+[ ! -d $sdir/english ] \
+    && echo Expecting directory $sdir/english to be present && exit 1;
+    [ -d $tdir/2000_hub5_eng_eval_tr ] \
+        && tdir=$tdir/2000_hub5_eng_eval_tr
+            [ ! -d $tdir/reference ] \
+                && echo Expecting directory $tdir/reference to be present && exit 1;
+
+
+            dir=data/local/eval2000
+            mkdir -p $dir
+
+            find $sdir/english -iname '*.sph' | sort > $dir/sph.flist
+            sed -e 's?.*/??' -e 's?.sph??' $dir/sph.flist | paste - $dir/sph.flist \
+                > $dir/sph.scp
+
+            sph2pipe=sph2pipe
+            [ ! -x $sph2pipe ] \
+                && echo "Could not execute the sph2pipe program at $sph2pipe" && exit 1;
+
+            awk -v sph2pipe=$sph2pipe '{
+            printf("%s-A %s -f wav -p -c 1 %s |\n", $1, sph2pipe, $2);
+            printf("%s-B %s -f wav -p -c 2 %s |\n", $1, sph2pipe, $2);
+        }' < $dir/sph.scp | sort > $dir/wav.scp || exit 1;
+    #side A - channel 1, side B - channel 2
+
+# Get segments file...
+# segments file format is: utt-id side-id start-time end-time, e.g.:
+# sw02001-A_000098-001156 sw02001-A 0.98 11.56
+pem=$sdir/english/hub5e_00.pem
+[ ! -f $pem ] && echo "$0: No such file $pem" && exit 1;
+# pem file has lines like:
+# en_4156 A unknown_speaker 301.85 302.48
+
+# we ignore the warnings below for now, although they seem to indicate some problems
+# with the data.
+grep -v ';;' $pem \
+    | awk '{
+    spk=$1"-"$2;
+    utt=sprintf("%s_%06d-%06d",spk,$4*100,$5*100);
+    print utt,spk,$4,$5;}' \
+        | sort -u | local/extend_segments.pl 0.1 > $dir/segments
+
+# stm file has lines like:
+# en_4156 A en_4156_A 357.64 359.64 <O,en,F,en-F>  HE IS A POLICE OFFICER
+# TODO(arnab): We should really be lowercasing this since the Edinburgh
+# recipe uses lowercase. This is not used in the actual scoring.
+grep -v ';;' $tdir/reference/hub5e00.english.000405.stm \
+    | awk '{
+    spk=$1"-"$2;
+    utt=sprintf("%s_%06d-%06d",spk,$4*100,$5*100);
+    printf utt; for(n=7;n<=NF;n++) printf(" %s", $n); print ""; }' \
+        | sort > $dir/text.all
+
+# We'll use the stm file for sclite scoring.  There seem to be various errors
+# in the stm file that upset hubscr.pl, and we fix them here.
+sed -e 's:((:(:' -e 's:<B_ASIDE>::g' -e 's:<E_ASIDE>::g' \
+    $tdir/reference/hub5e00.english.000405.stm >  $dir/stm
+    cp $tdir/reference/en20000405_hub5.glm  $dir/glm
+
+# next line uses command substitution
+# Just checking that the segments are the same in pem vs. stm.
+! cmp <(awk '{print $1}' $dir/text.all) <(awk '{print $1}' $dir/segments) && \
+    echo "Segments from pem file and stm file do not match." && exit 1;
+
+grep -v IGNORE_TIME_SEGMENT_ $dir/text.all > $dir/text
+
+# create an utt2spk file that assumes each conversation side is
+# a separate speaker.
+awk '{print $1,$2;}' $dir/segments > $dir/utt2spk
+utils/utt2spk_to_spk2utt.pl $dir/utt2spk > $dir/spk2utt
+
+# cp $dir/segments $dir/segments.tmp
+# awk '{x=$3-0.05; if (x<0.0) x=0.0; y=$4+0.05; print $1, $2, x, y; }' \
+#   $dir/segments.tmp > $dir/segments
+
+awk '{print $1}' $dir/wav.scp \
+    | perl -ane '$_ =~ m:^(\S+)-([AB])$: || die "bad label $_";
+    print "$1-$2 $1 $2\n"; ' \
+        > $dir/reco2file_and_channel || exit 1;
+
+
+    echo Data preparation and formatting completed for Eval 2000
+    echo "(but not MFCC extraction)"
+
+    utils/fix_data_dir.sh $dir
+
+    if [ $(wc -l < $dir/wav.scp) -ne 80 ]; then
+        echo "$0: error: expected 80 lines in wav.scp, got $(wc -l < $dir/wav.scp)"
+        exit 1;
+    fi

egs/swbd/ASR/local/extend_segments.pl

@@ -0,0 +1,99 @@
+#!/usr/bin/env perl
+use warnings; #sed replacement for -w perl parameter
+
+if (@ARGV != 1 || !($ARGV[0] =~ m/^-?\d+\.?\d*$/ &&  $ARGV[0] >= 0)) {
+  print STDERR "Usage: extend_segments.pl time-in-seconds <segments >segments.extended \n" .
+       "e.g. extend_segments.pl 0.25 <segments.1 >segments.2\n" .
+       "This command modifies a segments file, with lines like\n" .
+       " <utterance-id> <recording-id> <start-time> <end-time>\n" .
+       "by extending the beginning and end of each segment by a certain\n" .
+       "length of time.  This script makes sure the output segments do not\n" .
+       "overlap as a result of this time-extension, and that there are no\n" .
+       "negative times in the output.\n";
+  exit 1;
+}
+
+$extend = $ARGV[0];
+
+@all_lines = ();
+
+while (<STDIN>) {
+  chop;
+  @A = split(" ", $_);
+  if (@A != 4) {
+    die "invalid line in segments file: $_";
+  }
+  $line = @all_lines;  # current number of lines.
+  ($utt_id, $reco_id, $start_time, $end_time) = @A;
+
+  push @all_lines, [ $utt_id, $reco_id, $start_time, $end_time ]; # anonymous array.
+  if (! defined $lines_for_reco{$reco_id}) {
+    $lines_for_reco{$reco_id} = [ ];  # push new anonymous array.
+  }
+  push @{$lines_for_reco{$reco_id}}, $line;
+}
+
+foreach $reco_id (keys %lines_for_reco) {
+  $ref = $lines_for_reco{$reco_id};
+  @line_numbers = sort { ${$all_lines[$a]}[2] <=> ${$all_lines[$b]}[2] } @$ref;
+
+
+  {
+    # handle start of earliest segment as a special case.
+    $l0 = $line_numbers[0];
+    $tstart = ${$all_lines[$l0]}[2] - $extend;
+    if ($tstart < 0.0) { $tstart = 0.0; }
+    ${$all_lines[$l0]}[2] = $tstart;
+  }
+  {
+    # handle end of latest segment as a special case.
+    $lN = $line_numbers[$#line_numbers];
+    $tend = ${$all_lines[$lN]}[3] + $extend;
+    ${$all_lines[$lN]}[3] = $tend;
+  }
+  for ($i = 0; $i < $#line_numbers; $i++) {
+    $ln = $line_numbers[$i];
+    $ln1 = $line_numbers[$i+1];
+    $tend = ${$all_lines[$ln]}[3]; # end of earlier segment.
+    $tstart = ${$all_lines[$ln1]}[2]; # start of later segment.
+    if ($tend > $tstart) {
+      $utt1 = ${$all_lines[$ln]}[0];
+      $utt2 = ${$all_lines[$ln1]}[0];
+      print STDERR "Warning: for utterances $utt1 and $utt2, segments " .
+        "already overlap; leaving these times unchanged.\n";
+    } else {
+      $my_extend = $extend;
+      $max_extend =  0.5 * ($tstart - $tend);
+      if ($my_extend > $max_extend) { $my_extend = $max_extend; }
+      $tend += $my_extend;
+      $tstart -= $my_extend;
+      ${$all_lines[$ln]}[3] = $tend;
+      ${$all_lines[$ln1]}[2] = $tstart;
+    }
+  }
+}
+
+# leave the numbering of the lines unchanged.
+for ($l = 0; $l < @all_lines; $l++) {
+  $ref = $all_lines[$l];
+  ($utt_id, $reco_id, $start_time, $end_time) = @$ref;
+  printf("%s %s %.2f %.2f\n", $utt_id, $reco_id, $start_time, $end_time);
+}
+
+__END__
+
+# testing below.
+
+# ( echo a1 A 0 1; echo a2 A 3 4; echo b1 B 0 1; echo b2 B 2 3 ) | local/extend_segments.pl 1.0
+a1 A 0.00 2.00
+a2 A 2.00 5.00
+b1 B 0.00 1.50
+b2 B 1.50 4.00
+# ( echo a1 A 0 2; echo a2 A 1 3 ) | local/extend_segments.pl 1.0
+Warning: for utterances a1 and a2, segments already overlap; leaving these times unchanged.
+a1 A 0.00 2.00
+a2 A 1.00 4.00
+# ( echo a1 A 0 2; echo a2 A 5 6; echo a3 A 3 4 ) | local/extend_segments.pl 1.0
+a1 A 0.00 2.50
+a2 A 4.50 7.00
+a3 A 2.50 4.50

egs/swbd/ASR/local/filter_cuts.py

@@ -0,0 +1,160 @@
+#!/usr/bin/env python3
+# Copyright    2022  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+This script removes short and long utterances from a cutset.
+
+Caution:
+  You may need to tune the thresholds for your own dataset.
+
+Usage example:
+
+  python3 ./local/filter_cuts.py \
+    --bpe-model data/lang_bpe_500/bpe.model \
+    --in-cuts data/fbank/librispeech_cuts_test-clean.jsonl.gz \
+    --out-cuts data/fbank-filtered/librispeech_cuts_test-clean.jsonl.gz
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import sentencepiece as spm
+from lhotse import CutSet, load_manifest_lazy
+from lhotse.cut import Cut
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--bpe-model",
+        type=Path,
+        help="Path to the bpe.model",
+    )
+
+    parser.add_argument(
+        "--in-cuts",
+        type=Path,
+        help="Path to the input cutset",
+    )
+
+    parser.add_argument(
+        "--out-cuts",
+        type=Path,
+        help="Path to the output cutset",
+    )
+
+    return parser.parse_args()
+
+
+def filter_cuts(cut_set: CutSet, sp: spm.SentencePieceProcessor):
+    total = 0  # number of total utterances before removal
+    removed = 0  # number of removed utterances
+
+    def remove_short_and_long_utterances(c: Cut):
+        """Return False to exclude the input cut"""
+        nonlocal removed, total
+        # Keep only utterances with duration between 1 second and 20 seconds
+        #
+        # Caution: There is a reason to select 20.0 here. Please see
+        # ./display_manifest_statistics.py
+        #
+        # You should use ./display_manifest_statistics.py to get
+        # an utterance duration distribution for your dataset to select
+        # the threshold
+        total += 1
+        if c.duration < 1.0 or c.duration > 20.0:
+            logging.warning(
+                f"Exclude cut with ID {c.id} from training. Duration: {c.duration}"
+            )
+            removed += 1
+            return False
+
+        # In pruned RNN-T, we require that T >= S
+        # where T is the number of feature frames after subsampling
+        # and S is the number of tokens in the utterance
+
+        # In ./pruned_transducer_stateless2/conformer.py, the
+        # conv module uses the following expression
+        # for subsampling
+        if c.num_frames is None:
+            num_frames = c.duration * 100  # approximate
+        else:
+            num_frames = c.num_frames
+
+        T = ((num_frames - 1) // 2 - 1) // 2
+        # Note: for ./lstm_transducer_stateless/lstm.py, the formula is
+        #  T = ((num_frames - 3) // 2 - 1) // 2
+
+        # Note: for ./pruned_transducer_stateless7/zipformer.py, the formula is
+        # T = ((num_frames - 7) // 2 + 1) // 2
+
+        tokens = sp.encode(c.supervisions[0].text, out_type=str)
+
+        if T < len(tokens):
+            logging.warning(
+                f"Exclude cut with ID {c.id} from training. "
+                f"Number of frames (before subsampling): {c.num_frames}. "
+                f"Number of frames (after subsampling): {T}. "
+                f"Text: {c.supervisions[0].text}. "
+                f"Tokens: {tokens}. "
+                f"Number of tokens: {len(tokens)}"
+            )
+            removed += 1
+            return False
+
+        return True
+
+    # We use to_eager() here so that we can print out the value of total
+    # and removed below.
+    ans = cut_set.filter(remove_short_and_long_utterances).to_eager()
+    ratio = removed / total * 100
+    logging.info(
+        f"Removed {removed} cuts from {total} cuts. {ratio:.3f}% data is removed."
+    )
+    return ans
+
+
+def main():
+    args = get_args()
+    logging.info(vars(args))
+
+    if args.out_cuts.is_file():
+        logging.info(f"{args.out_cuts} already exists - skipping")
+        return
+
+    assert args.in_cuts.is_file(), f"{args.in_cuts} does not exist"
+    assert args.bpe_model.is_file(), f"{args.bpe_model} does not exist"
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(str(args.bpe_model))
+
+    cut_set = load_manifest_lazy(args.in_cuts)
+    assert isinstance(cut_set, CutSet)
+
+    cut_set = filter_cuts(cut_set, sp)
+    logging.info(f"Saving to {args.out_cuts}")
+    args.out_cuts.parent.mkdir(parents=True, exist_ok=True)
+    cut_set.to_file(args.out_cuts)
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/swbd/ASR/local/filter_empty_text.py

@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+# Copyright    2023  The Chinese University of Hong Kong (author: Zengrui Jin)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+from pathlib import Path
+import logging
+from typing import List
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--kaldi-data-dir",
+        type=Path,
+        required=True,
+        help="Path to the kaldi data dir",
+    )
+
+    return parser.parse_args()
+
+
+def load_segments(path: Path):
+    segments = {}
+    with open(path, "r") as f:
+        lines = f.readlines()
+        for line in lines:
+            line = line.strip()
+            utt_id, rec_id, start, end = line.split()
+            segments[utt_id] = line
+    return segments
+
+
+def filter_text(path: Path):
+    with open(path, "r") as f:
+        lines = f.readlines()
+        return list(filter(lambda x: len(x.strip().split()) > 1, lines))
+
+
+def write_segments(path: Path, texts: List[str]):
+    with open(path, "w") as f:
+        f.writelines(texts)
+
+
+def main():
+    args = get_args()
+    orig_text_dict = filter_text(args.kaldi_data_dir / "text")
+    write_segments(args.kaldi_data_dir / "text", orig_text_dict)
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()
+
+    logging.info("Empty lines filtered")

egs/swbd/ASR/local/format_acronyms_dict.py

@@ -0,0 +1,118 @@
+#!/usr/bin/env python3
+
+# Copyright 2015  Minhua Wu
+# Apache 2.0
+
+# convert acronyms in swbd dict to fisher convention
+# IBM to i._b._m.
+# BBC to b._b._c.
+# BBCs to b._b._c.s
+# BBC's to b._b._c.'s
+
+import argparse
+import re
+
+__author__ = "Minhua Wu"
+
+parser = argparse.ArgumentParser(description="format acronyms to a._b._c.")
+parser.add_argument("-i", "--input", help="Input lexicon", required=True)
+parser.add_argument("-o", "--output", help="Output lexicon", required=True)
+parser.add_argument(
+    "-L", "--Letter", help="Input single letter pronunciation", required=True
+)
+parser.add_argument("-M", "--Map", help="Output acronyms mapping", required=True)
+args = parser.parse_args()
+
+
+fin_lex = open(args.input, "r")
+fin_Letter = open(args.Letter, "r")
+fout_lex = open(args.output, "w")
+fout_map = open(args.Map, "w")
+
+# Initialise single letter dictionary
+dict_letter = {}
+for single_letter_lex in fin_Letter:
+    items = single_letter_lex.split()
+    dict_letter[items[0]] = single_letter_lex[len(items[0]) + 1 :].strip()
+fin_Letter.close()
+# print dict_letter
+
+for lex in fin_lex:
+    items = lex.split()
+    word = items[0]
+    lexicon = lex[len(items[0]) + 1 :].strip()
+    # find acronyms from words with only letters and '
+    pre_match = re.match(r"^[A-Za-z]+$|^[A-Za-z]+\'s$|^[A-Za-z]+s$", word)
+    if pre_match:
+        # find if words in the form of xxx's is acronym
+        if word[-2:] == "'s" and (lexicon[-1] == "s" or lexicon[-1] == "z"):
+            actual_word = word[:-2]
+            actual_lexicon = lexicon[:-2]
+            acronym_lexicon = ""
+            for w in actual_word:
+                acronym_lexicon = acronym_lexicon + dict_letter[w.upper()] + " "
+            if acronym_lexicon.strip() == actual_lexicon:
+                acronym_mapped = ""
+                acronym_mapped_back = ""
+                for w in actual_word[:-1]:
+                    acronym_mapped = acronym_mapped + w.lower() + "._"
+                    acronym_mapped_back = acronym_mapped_back + w.lower() + " "
+                acronym_mapped = acronym_mapped + actual_word[-1].lower() + ".'s"
+                acronym_mapped_back = (
+                    acronym_mapped_back + actual_word[-1].lower() + "'s"
+                )
+                fout_map.write(
+                    word + "\t" + acronym_mapped + "\t" + acronym_mapped_back + "\n"
+                )
+                fout_lex.write(acronym_mapped + " " + lexicon + "\n")
+            else:
+                fout_lex.write(lex)
+
+        # find if words in the form of xxxs is acronym
+        elif word[-1] == "s" and (lexicon[-1] == "s" or lexicon[-1] == "z"):
+            actual_word = word[:-1]
+            actual_lexicon = lexicon[:-2]
+            acronym_lexicon = ""
+            for w in actual_word:
+                acronym_lexicon = acronym_lexicon + dict_letter[w.upper()] + " "
+            if acronym_lexicon.strip() == actual_lexicon:
+                acronym_mapped = ""
+                acronym_mapped_back = ""
+                for w in actual_word[:-1]:
+                    acronym_mapped = acronym_mapped + w.lower() + "._"
+                    acronym_mapped_back = acronym_mapped_back + w.lower() + " "
+                acronym_mapped = acronym_mapped + actual_word[-1].lower() + ".s"
+                acronym_mapped_back = (
+                    acronym_mapped_back + actual_word[-1].lower() + "'s"
+                )
+                fout_map.write(
+                    word + "\t" + acronym_mapped + "\t" + acronym_mapped_back + "\n"
+                )
+                fout_lex.write(acronym_mapped + " " + lexicon + "\n")
+            else:
+                fout_lex.write(lex)
+
+        # find if words in the form of xxx (not ended with 's or s) is acronym
+        elif word.find("'") == -1 and word[-1] != "s":
+            acronym_lexicon = ""
+            for w in word:
+                acronym_lexicon = acronym_lexicon + dict_letter[w.upper()] + " "
+            if acronym_lexicon.strip() == lexicon:
+                acronym_mapped = ""
+                acronym_mapped_back = ""
+                for w in word[:-1]:
+                    acronym_mapped = acronym_mapped + w.lower() + "._"
+                    acronym_mapped_back = acronym_mapped_back + w.lower() + " "
+                acronym_mapped = acronym_mapped + word[-1].lower() + "."
+                acronym_mapped_back = acronym_mapped_back + word[-1].lower()
+                fout_map.write(
+                    word + "\t" + acronym_mapped + "\t" + acronym_mapped_back + "\n"
+                )
+                fout_lex.write(acronym_mapped + " " + lexicon + "\n")
+            else:
+                fout_lex.write(lex)
+        else:
+            fout_lex.write(lex)
+
+    else:
+        fout_lex.write(lex)

egs/swbd/ASR/local/generate_unique_lexicon.py

@@ -0,0 +1,98 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+This file takes as input a lexicon.txt and output a new lexicon,
+in which each word has a unique pronunciation.
+
+The way to do this is to keep only the first pronunciation of a word
+in lexicon.txt.
+"""
+
+
+import argparse
+import logging
+from pathlib import Path
+from typing import List, Tuple
+
+from icefall.lexicon import read_lexicon, write_lexicon
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        It should contain a file lexicon.txt.
+        This file will generate a new file uniq_lexicon.txt
+        in it.
+        """,
+    )
+
+    return parser.parse_args()
+
+
+def filter_multiple_pronunications(
+    lexicon: List[Tuple[str, List[str]]]
+) -> List[Tuple[str, List[str]]]:
+    """Remove multiple pronunciations of words from a lexicon.
+
+    If a word has more than one pronunciation in the lexicon, only
+    the first one is kept, while other pronunciations are removed
+    from the lexicon.
+
+    Args:
+      lexicon:
+        The input lexicon, containing a list of (word, [p1, p2, ..., pn]),
+        where "p1, p2, ..., pn" are the pronunciations of the "word".
+    Returns:
+      Return a new lexicon where each word has a unique pronunciation.
+    """
+    seen = set()
+    ans = []
+
+    for word, tokens in lexicon:
+        if word in seen:
+            continue
+        seen.add(word)
+        ans.append((word, tokens))
+    return ans
+
+
+def main():
+    args = get_args()
+    lang_dir = Path(args.lang_dir)
+
+    lexicon_filename = lang_dir / "lexicon.txt"
+
+    in_lexicon = read_lexicon(lexicon_filename)
+
+    out_lexicon = filter_multiple_pronunications(in_lexicon)
+
+    write_lexicon(lang_dir / "uniq_lexicon.txt", out_lexicon)
+
+    logging.info(f"Number of entries in lexicon.txt: {len(in_lexicon)}")
+    logging.info(f"Number of entries in uniq_lexicon.txt: {len(out_lexicon)}")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/swbd/ASR/local/map_acronyms_transcripts.py

@@ -0,0 +1,60 @@
+#!/usr/bin/env python3
+
+# Copyright 2015  Minhua Wu
+# Apache 2.0
+
+# convert acronyms in swbd transcript to fisher convention
+# according to first two columns in the input acronyms mapping
+
+import argparse
+import re
+
+__author__ = "Minhua Wu"
+
+parser = argparse.ArgumentParser(description="format acronyms to a._b._c.")
+parser.add_argument("-i", "--input", help="Input transcripts", required=True)
+parser.add_argument("-o", "--output", help="Output transcripts", required=True)
+parser.add_argument("-M", "--Map", help="Input acronyms mapping", required=True)
+args = parser.parse_args()
+
+fin_map = open(args.Map, "r")
+dict_acronym = {}
+dict_acronym_noi = {}  # Mapping of acronyms without I, i
+for pair in fin_map:
+    items = pair.split("\t")
+    dict_acronym[items[0]] = items[1]
+    dict_acronym_noi[items[0]] = items[1]
+fin_map.close()
+del dict_acronym_noi["I"]
+del dict_acronym_noi["i"]
+
+
+fin_trans = open(args.input, "r")
+fout_trans = open(args.output, "w")
+for line in fin_trans:
+    items = line.split()
+    L = len(items)
+    # First pass mapping to map I as part of acronym
+    for i in range(L):
+        if items[i] == "I":
+            x = 0
+            while i - 1 - x >= 0 and re.match(r"^[A-Z]$", items[i - 1 - x]):
+                x += 1
+
+            y = 0
+            while i + 1 + y < L and re.match(r"^[A-Z]$", items[i + 1 + y]):
+                y += 1
+
+            if x + y > 0:
+                for bias in range(-x, y + 1):
+                    items[i + bias] = dict_acronym[items[i + bias]]
+
+    # Second pass mapping (not mapping 'i' and 'I')
+    for i in range(len(items)):
+        if items[i] in dict_acronym_noi.keys():
+            items[i] = dict_acronym_noi[items[i]]
+    sentence = " ".join(items[1:])
+    fout_trans.write(items[0] + " " + sentence.lower() + "\n")
+
+fin_trans.close()
+fout_trans.close()

egs/swbd/ASR/local/prepare_lang.py

@@ -0,0 +1,413 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This script takes as input a lexicon file "data/lang_phone/lexicon.txt"
+consisting of words and tokens (i.e., phones) and does the following:
+
+1. Add disambiguation symbols to the lexicon and generate lexicon_disambig.txt
+
+2. Generate tokens.txt, the token table mapping a token to a unique integer.
+
+3. Generate words.txt, the word table mapping a word to a unique integer.
+
+4. Generate L.pt, in k2 format. It can be loaded by
+
+        d = torch.load("L.pt")
+        lexicon = k2.Fsa.from_dict(d)
+
+5. Generate L_disambig.pt, in k2 format.
+"""
+import argparse
+import math
+from collections import defaultdict
+from pathlib import Path
+from typing import Any, Dict, List, Tuple
+
+import k2
+import torch
+
+from icefall.lexicon import read_lexicon, write_lexicon
+from icefall.utils import str2bool
+
+Lexicon = List[Tuple[str, List[str]]]
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        It should contain a file lexicon.txt.
+        Generated files by this script are saved into this directory.
+        """,
+    )
+
+    parser.add_argument(
+        "--debug",
+        type=str2bool,
+        default=False,
+        help="""True for debugging, which will generate
+        a visualization of the lexicon FST.
+
+        Caution: If your lexicon contains hundreds of thousands
+        of lines, please set it to False!
+        """,
+    )
+
+    return parser.parse_args()
+
+
+def write_mapping(filename: str, sym2id: Dict[str, int]) -> None:
+    """Write a symbol to ID mapping to a file.
+
+    Note:
+      No need to implement `read_mapping` as it can be done
+      through :func:`k2.SymbolTable.from_file`.
+
+    Args:
+      filename:
+        Filename to save the mapping.
+      sym2id:
+        A dict mapping symbols to IDs.
+    Returns:
+      Return None.
+    """
+    with open(filename, "w", encoding="utf-8") as f:
+        for sym, i in sym2id.items():
+            f.write(f"{sym} {i}\n")
+
+
+def get_tokens(lexicon: Lexicon) -> List[str]:
+    """Get tokens from a lexicon.
+
+    Args:
+      lexicon:
+        It is the return value of :func:`read_lexicon`.
+    Returns:
+      Return a list of unique tokens.
+    """
+    ans = set()
+    for _, tokens in lexicon:
+        ans.update(tokens)
+    sorted_ans = sorted(list(ans))
+    return sorted_ans
+
+
+def get_words(lexicon: Lexicon) -> List[str]:
+    """Get words from a lexicon.
+
+    Args:
+      lexicon:
+        It is the return value of :func:`read_lexicon`.
+    Returns:
+      Return a list of unique words.
+    """
+    ans = set()
+    for word, _ in lexicon:
+        ans.add(word)
+    sorted_ans = sorted(list(ans))
+    return sorted_ans
+
+
+def add_disambig_symbols(lexicon: Lexicon) -> Tuple[Lexicon, int]:
+    """It adds pseudo-token disambiguation symbols #1, #2 and so on
+    at the ends of tokens to ensure that all pronunciations are different,
+    and that none is a prefix of another.
+
+    See also add_lex_disambig.pl from kaldi.
+
+    Args:
+      lexicon:
+        It is returned by :func:`read_lexicon`.
+    Returns:
+      Return a tuple with two elements:
+
+        - The output lexicon with disambiguation symbols
+        - The ID of the max disambiguation symbol that appears
+          in the lexicon
+    """
+
+    # (1) Work out the count of each token-sequence in the
+    # lexicon.
+    count = defaultdict(int)
+    for _, tokens in lexicon:
+        count[" ".join(tokens)] += 1
+
+    # (2) For each left sub-sequence of each token-sequence, note down
+    # that it exists (for identifying prefixes of longer strings).
+    issubseq = defaultdict(int)
+    for _, tokens in lexicon:
+        tokens = tokens.copy()
+        tokens.pop()
+        while tokens:
+            issubseq[" ".join(tokens)] = 1
+            tokens.pop()
+
+    # (3) For each entry in the lexicon:
+    # if the token sequence is unique and is not a
+    # prefix of another word, no disambig symbol.
+    # Else output #1, or #2, #3, ... if the same token-seq
+    # has already been assigned a disambig symbol.
+    ans = []
+
+    # We start with #1 since #0 has its own purpose
+    first_allowed_disambig = 1
+    max_disambig = first_allowed_disambig - 1
+    last_used_disambig_symbol_of = defaultdict(int)
+
+    for word, tokens in lexicon:
+        tokenseq = " ".join(tokens)
+        assert tokenseq != ""
+        if issubseq[tokenseq] == 0 and count[tokenseq] == 1:
+            ans.append((word, tokens))
+            continue
+
+        cur_disambig = last_used_disambig_symbol_of[tokenseq]
+        if cur_disambig == 0:
+            cur_disambig = first_allowed_disambig
+        else:
+            cur_disambig += 1
+
+        if cur_disambig > max_disambig:
+            max_disambig = cur_disambig
+        last_used_disambig_symbol_of[tokenseq] = cur_disambig
+        tokenseq += f" #{cur_disambig}"
+        ans.append((word, tokenseq.split()))
+    return ans, max_disambig
+
+
+def generate_id_map(symbols: List[str]) -> Dict[str, int]:
+    """Generate ID maps, i.e., map a symbol to a unique ID.
+
+    Args:
+      symbols:
+        A list of unique symbols.
+    Returns:
+      A dict containing the mapping between symbols and IDs.
+    """
+    return {sym: i for i, sym in enumerate(symbols)}
+
+
+def add_self_loops(
+    arcs: List[List[Any]], disambig_token: int, disambig_word: int
+) -> List[List[Any]]:
+    """Adds self-loops to states of an FST to propagate disambiguation symbols
+    through it. They are added on each state with non-epsilon output symbols
+    on at least one arc out of the state.
+
+    See also fstaddselfloops.pl from Kaldi. One difference is that
+    Kaldi uses OpenFst style FSTs and it has multiple final states.
+    This function uses k2 style FSTs and it does not need to add self-loops
+    to the final state.
+
+    The input label of a self-loop is `disambig_token`, while the output
+    label is `disambig_word`.
+
+    Args:
+      arcs:
+        A list-of-list. The sublist contains
+        `[src_state, dest_state, label, aux_label, score]`
+      disambig_token:
+        It is the token ID of the symbol `#0`.
+      disambig_word:
+        It is the word ID of the symbol `#0`.
+
+    Return:
+      Return new `arcs` containing self-loops.
+    """
+    states_needs_self_loops = set()
+    for arc in arcs:
+        src, dst, ilabel, olabel, score = arc
+        if olabel != 0:
+            states_needs_self_loops.add(src)
+
+    ans = []
+    for s in states_needs_self_loops:
+        ans.append([s, s, disambig_token, disambig_word, 0])
+
+    return arcs + ans
+
+
+def lexicon_to_fst(
+    lexicon: Lexicon,
+    token2id: Dict[str, int],
+    word2id: Dict[str, int],
+    sil_token: str = "sil",
+    sil_prob: float = 0.5,
+    need_self_loops: bool = False,
+) -> k2.Fsa:
+    """Convert a lexicon to an FST (in k2 format) with optional silence at
+    the beginning and end of each word.
+
+    Args:
+      lexicon:
+        The input lexicon. See also :func:`read_lexicon`
+      token2id:
+        A dict mapping tokens to IDs.
+      word2id:
+        A dict mapping words to IDs.
+      sil_token:
+        The silence token.
+      sil_prob:
+        The probability for adding a silence at the beginning and end
+        of the word.
+      need_self_loops:
+        If True, add self-loop to states with non-epsilon output symbols
+        on at least one arc out of the state. The input label for this
+        self loop is `token2id["#0"]` and the output label is `word2id["#0"]`.
+    Returns:
+      Return an instance of `k2.Fsa` representing the given lexicon.
+    """
+    assert sil_prob > 0.0 and sil_prob < 1.0
+    # CAUTION: we use score, i.e, negative cost.
+    sil_score = math.log(sil_prob)
+    no_sil_score = math.log(1.0 - sil_prob)
+
+    start_state = 0
+    loop_state = 1  # words enter and leave from here
+    sil_state = 2  # words terminate here when followed by silence; this state
+    # has a silence transition to loop_state.
+    next_state = 3  # the next un-allocated state, will be incremented as we go.
+    arcs = []
+
+    assert token2id["<eps>"] == 0
+    assert word2id["<eps>"] == 0
+
+    eps = 0
+
+    sil_token = token2id[sil_token]
+
+    arcs.append([start_state, loop_state, eps, eps, no_sil_score])
+    arcs.append([start_state, sil_state, eps, eps, sil_score])
+    arcs.append([sil_state, loop_state, sil_token, eps, 0])
+
+    for word, tokens in lexicon:
+        assert len(tokens) > 0, f"{word} has no pronunciations"
+        cur_state = loop_state
+
+        word = word2id[word]
+        tokens = [token2id[i] for i in tokens]
+
+        for i in range(len(tokens) - 1):
+            w = word if i == 0 else eps
+            arcs.append([cur_state, next_state, tokens[i], w, 0])
+
+            cur_state = next_state
+            next_state += 1
+
+        # now for the last token of this word
+        # It has two out-going arcs, one to the loop state,
+        # the other one to the sil_state.
+        i = len(tokens) - 1
+        w = word if i == 0 else eps
+        arcs.append([cur_state, loop_state, tokens[i], w, no_sil_score])
+        arcs.append([cur_state, sil_state, tokens[i], w, sil_score])
+
+    if need_self_loops:
+        disambig_token = token2id["#0"]
+        disambig_word = word2id["#0"]
+        arcs = add_self_loops(
+            arcs,
+            disambig_token=disambig_token,
+            disambig_word=disambig_word,
+        )
+
+    final_state = next_state
+    arcs.append([loop_state, final_state, -1, -1, 0])
+    arcs.append([final_state])
+
+    arcs = sorted(arcs, key=lambda arc: arc[0])
+    arcs = [[str(i) for i in arc] for arc in arcs]
+    arcs = [" ".join(arc) for arc in arcs]
+    arcs = "\n".join(arcs)
+
+    fsa = k2.Fsa.from_str(arcs, acceptor=False)
+    return fsa
+
+
+def main():
+    args = get_args()
+    lang_dir = Path(args.lang_dir)
+    lexicon_filename = lang_dir / "lexicon.txt"
+    sil_token = "sil"
+    sil_prob = 0.5
+
+    lexicon = read_lexicon(lexicon_filename)
+    tokens = get_tokens(lexicon)
+    words = get_words(lexicon)
+
+    lexicon_disambig, max_disambig = add_disambig_symbols(lexicon)
+
+    for i in range(max_disambig + 1):
+        disambig = f"#{i}"
+        assert disambig not in tokens
+        tokens.append(f"#{i}")
+
+    assert "<eps>" not in tokens
+    tokens = ["<eps>"] + tokens
+
+    assert "<eps>" not in words
+    assert "#0" not in words
+    assert "<s>" not in words
+    assert "</s>" not in words
+
+    words = ["<eps>"] + words + ["#0", "<s>", "</s>"]
+
+    token2id = generate_id_map(tokens)
+    word2id = generate_id_map(words)
+
+    write_mapping(lang_dir / "tokens.txt", token2id)
+    write_mapping(lang_dir / "words.txt", word2id)
+    write_lexicon(lang_dir / "lexicon_disambig.txt", lexicon_disambig)
+
+    L = lexicon_to_fst(
+        lexicon,
+        token2id=token2id,
+        word2id=word2id,
+        sil_token=sil_token,
+        sil_prob=sil_prob,
+    )
+
+    L_disambig = lexicon_to_fst(
+        lexicon_disambig,
+        token2id=token2id,
+        word2id=word2id,
+        sil_token=sil_token,
+        sil_prob=sil_prob,
+        need_self_loops=True,
+    )
+    torch.save(L.as_dict(), lang_dir / "L.pt")
+    torch.save(L_disambig.as_dict(), lang_dir / "L_disambig.pt")
+
+    if args.debug:
+        labels_sym = k2.SymbolTable.from_file(lang_dir / "tokens.txt")
+        aux_labels_sym = k2.SymbolTable.from_file(lang_dir / "words.txt")
+
+        L.labels_sym = labels_sym
+        L.aux_labels_sym = aux_labels_sym
+        L.draw(f"{lang_dir / 'L.svg'}", title="L.pt")
+
+        L_disambig.labels_sym = labels_sym
+        L_disambig.aux_labels_sym = aux_labels_sym
+        L_disambig.draw(f"{lang_dir / 'L_disambig.svg'}", title="L_disambig.pt")
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/local/prepare_lang_bpe.py

@@ -0,0 +1,277 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+# Copyright (c)  2021  Xiaomi Corporation (authors: Fangjun Kuang)
+
+"""
+
+This script takes as input `lang_dir`, which should contain::
+
+    - lang_dir/bpe.model,
+    - lang_dir/words.txt
+
+and generates the following files in the directory `lang_dir`:
+
+    - lexicon.txt
+    - lexicon_disambig.txt
+    - L.pt
+    - L_disambig.pt
+    - tokens.txt
+"""
+
+import argparse
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+from prepare_lang import (
+    Lexicon,
+    add_disambig_symbols,
+    add_self_loops,
+    write_lexicon,
+    write_mapping,
+)
+
+from icefall.utils import str2bool
+
+
+def lexicon_to_fst_no_sil(
+    lexicon: Lexicon,
+    token2id: Dict[str, int],
+    word2id: Dict[str, int],
+    need_self_loops: bool = False,
+) -> k2.Fsa:
+    """Convert a lexicon to an FST (in k2 format).
+
+    Args:
+      lexicon:
+        The input lexicon. See also :func:`read_lexicon`
+      token2id:
+        A dict mapping tokens to IDs.
+      word2id:
+        A dict mapping words to IDs.
+      need_self_loops:
+        If True, add self-loop to states with non-epsilon output symbols
+        on at least one arc out of the state. The input label for this
+        self loop is `token2id["#0"]` and the output label is `word2id["#0"]`.
+    Returns:
+      Return an instance of `k2.Fsa` representing the given lexicon.
+    """
+    loop_state = 0  # words enter and leave from here
+    next_state = 1  # the next un-allocated state, will be incremented as we go
+
+    arcs = []
+
+    # The blank symbol <blk> is defined in local/train_bpe_model.py
+    assert token2id["<blk>"] == 0
+    assert word2id["<eps>"] == 0
+
+    eps = 0
+
+    for word, pieces in lexicon:
+        assert len(pieces) > 0, f"{word} has no pronunciations"
+        cur_state = loop_state
+
+        word = word2id[word]
+        pieces = [token2id[i] for i in pieces]
+
+        for i in range(len(pieces) - 1):
+            w = word if i == 0 else eps
+            arcs.append([cur_state, next_state, pieces[i], w, 0])
+
+            cur_state = next_state
+            next_state += 1
+
+        # now for the last piece of this word
+        i = len(pieces) - 1
+        w = word if i == 0 else eps
+        arcs.append([cur_state, loop_state, pieces[i], w, 0])
+
+    if need_self_loops:
+        disambig_token = token2id["#0"]
+        disambig_word = word2id["#0"]
+        arcs = add_self_loops(
+            arcs,
+            disambig_token=disambig_token,
+            disambig_word=disambig_word,
+        )
+
+    final_state = next_state
+    arcs.append([loop_state, final_state, -1, -1, 0])
+    arcs.append([final_state])
+
+    arcs = sorted(arcs, key=lambda arc: arc[0])
+    arcs = [[str(i) for i in arc] for arc in arcs]
+    arcs = [" ".join(arc) for arc in arcs]
+    arcs = "\n".join(arcs)
+
+    fsa = k2.Fsa.from_str(arcs, acceptor=False)
+    return fsa
+
+
+def generate_lexicon(
+    model_file: str, words: List[str], oov: str
+) -> Tuple[Lexicon, Dict[str, int]]:
+    """Generate a lexicon from a BPE model.
+
+    Args:
+      model_file:
+        Path to a sentencepiece model.
+      words:
+        A list of strings representing words.
+      oov:
+        The out of vocabulary word in lexicon.
+    Returns:
+      Return a tuple with two elements:
+        - A dict whose keys are words and values are the corresponding
+          word pieces.
+        - A dict representing the token symbol, mapping from tokens to IDs.
+    """
+    sp = spm.SentencePieceProcessor()
+    sp.load(str(model_file))
+
+    # Convert word to word piece IDs instead of word piece strings
+    # to avoid OOV tokens.
+    words_pieces_ids: List[List[int]] = sp.encode(words, out_type=int)
+
+    # Now convert word piece IDs back to word piece strings.
+    words_pieces: List[List[str]] = [sp.id_to_piece(ids) for ids in words_pieces_ids]
+
+    lexicon = []
+    for word, pieces in zip(words, words_pieces):
+        lexicon.append((word, pieces))
+
+    lexicon.append((oov, ["▁", sp.id_to_piece(sp.unk_id())]))
+
+    token2id: Dict[str, int] = {sp.id_to_piece(i): i for i in range(sp.vocab_size())}
+
+    return lexicon, token2id
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        It should contain the bpe.model and words.txt
+        """,
+    )
+
+    parser.add_argument(
+        "--oov",
+        type=str,
+        default="<unk>",
+        help="The out of vocabulary word in lexicon.",
+    )
+
+    parser.add_argument(
+        "--debug",
+        type=str2bool,
+        default=False,
+        help="""True for debugging, which will generate
+        a visualization of the lexicon FST.
+
+        Caution: If your lexicon contains hundreds of thousands
+        of lines, please set it to False!
+
+        See "test/test_bpe_lexicon.py" for usage.
+        """,
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    lang_dir = Path(args.lang_dir)
+    model_file = lang_dir / "bpe.model"
+
+    word_sym_table = k2.SymbolTable.from_file(lang_dir / "words.txt")
+
+    words = word_sym_table.symbols
+
+    excluded = [
+        "<eps>",
+        "!sil",
+        "<spoken_noise>",
+        args.oov,
+        "#0",
+        "<s>",
+        "</s>",
+        "[vocalized-noise]",
+        "[noise]",
+        "[laughter]",
+    ]
+
+    for w in excluded:
+        if w in words:
+            words.remove(w)
+
+    lexicon, token_sym_table = generate_lexicon(model_file, words, args.oov)
+
+    lexicon_disambig, max_disambig = add_disambig_symbols(lexicon)
+
+    next_token_id = max(token_sym_table.values()) + 1
+    for i in range(max_disambig + 1):
+        disambig = f"#{i}"
+        assert disambig not in token_sym_table
+        token_sym_table[disambig] = next_token_id
+        next_token_id += 1
+
+    word_sym_table.add("#0")
+    word_sym_table.add("<s>")
+    word_sym_table.add("</s>")
+
+    write_mapping(lang_dir / "tokens.txt", token_sym_table)
+
+    write_lexicon(lang_dir / "lexicon.txt", lexicon)
+    write_lexicon(lang_dir / "lexicon_disambig.txt", lexicon_disambig)
+
+    L = lexicon_to_fst_no_sil(
+        lexicon,
+        token2id=token_sym_table,
+        word2id=word_sym_table,
+    )
+
+    L_disambig = lexicon_to_fst_no_sil(
+        lexicon_disambig,
+        token2id=token_sym_table,
+        word2id=word_sym_table,
+        need_self_loops=True,
+    )
+    torch.save(L.as_dict(), lang_dir / "L.pt")
+    torch.save(L_disambig.as_dict(), lang_dir / "L_disambig.pt")
+
+    if args.debug:
+        labels_sym = k2.SymbolTable.from_file(lang_dir / "tokens.txt")
+        aux_labels_sym = k2.SymbolTable.from_file(lang_dir / "words.txt")
+
+        L.labels_sym = labels_sym
+        L.aux_labels_sym = aux_labels_sym
+        L.draw(f"{lang_dir / 'L.svg'}", title="L.pt")
+
+        L_disambig.labels_sym = labels_sym
+        L_disambig.aux_labels_sym = aux_labels_sym
+        L_disambig.draw(f"{lang_dir / 'L_disambig.svg'}", title="L_disambig.pt")
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/local/prepare_lm_training_data.py

@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+
+# Copyright (c)  2021  Xiaomi Corporation (authors: Daniel Povey
+#                                                   Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+This script takes a `bpe.model` and a text file such as
+./download/lm/librispeech-lm-norm.txt
+and outputs the LM training data to a supplied directory such
+as data/lm_training_bpe_500.  The format is as follows:
+
+It creates a PyTorch archive (.pt file), say data/lm_training.pt, which is a
+representation of a dict with the following format:
+
+  'words' -> a k2.RaggedTensor of two axes [word][token] with dtype torch.int32
+             containing the BPE representations of each word, indexed by
+             integer word ID. (These integer word IDS are present in
+             'lm_data').  The sentencepiece object can be used to turn the
+             words and BPE units into string form.
+  'sentences' -> a k2.RaggedTensor of two axes [sentence][word] with dtype
+            torch.int32 containing all the sentences, as word-ids (we don't
+            output the string form of this directly but it can be worked out
+            together with 'words' and the bpe.model).
+  'sentence_lengths' -> a 1-D torch.Tensor of dtype torch.int32, containing
+            number of BPE tokens of each sentence.
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import k2
+import sentencepiece as spm
+import torch
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        help="Input BPE model, e.g. data/bpe_500/bpe.model",
+    )
+    parser.add_argument(
+        "--lm-data",
+        type=str,
+        help="""Input LM training data as text, e.g.
+        download/pb.train.txt""",
+    )
+    parser.add_argument(
+        "--lm-archive",
+        type=str,
+        help="""Path to output archive, e.g. data/bpe_500/lm_data.pt;
+        look at the source of this script to see the format.""",
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+
+    if Path(args.lm_archive).exists():
+        logging.warning(f"{args.lm_archive} exists - skipping")
+        return
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(args.bpe_model)
+
+    # word2index is a dictionary from words to integer ids.  No need to reserve
+    # space for epsilon, etc.; the words are just used as a convenient way to
+    # compress the sequences of BPE pieces.
+    word2index = dict()
+
+    word2bpe = []  # Will be a list-of-list-of-int, representing BPE pieces.
+    sentences = []  # Will be a list-of-list-of-int, representing word-ids.
+
+    if "librispeech-lm-norm" in args.lm_data:
+        num_lines_in_total = 40418261.0
+        step = 5000000
+    elif "valid" in args.lm_data:
+        num_lines_in_total = 5567.0
+        step = 3000
+    elif "test" in args.lm_data:
+        num_lines_in_total = 5559.0
+        step = 3000
+    else:
+        num_lines_in_total = None
+        step = None
+
+    processed = 0
+
+    with open(args.lm_data) as f:
+        while True:
+            line = f.readline()
+            if line == "":
+                break
+
+            if step and processed % step == 0:
+                logging.info(
+                    f"Processed number of lines: {processed} "
+                    f"({processed/num_lines_in_total*100: .3f}%)"
+                )
+            processed += 1
+
+            line_words = line.split()
+            for w in line_words:
+                if w not in word2index:
+                    w_bpe = sp.encode(w)
+                    word2index[w] = len(word2bpe)
+                    word2bpe.append(w_bpe)
+            sentences.append([word2index[w] for w in line_words])
+
+    logging.info("Constructing ragged tensors")
+    words = k2.ragged.RaggedTensor(word2bpe)
+    sentences = k2.ragged.RaggedTensor(sentences)
+
+    output = dict(words=words, sentences=sentences)
+
+    num_sentences = sentences.dim0
+    logging.info(f"Computing sentence lengths, num_sentences: {num_sentences}")
+    sentence_lengths = [0] * num_sentences
+    for i in range(num_sentences):
+        if step and i % step == 0:
+            logging.info(
+                f"Processed number of lines: {i} ({i/num_sentences*100: .3f}%)"
+            )
+
+        word_ids = sentences[i]
+
+        # NOTE: If word_ids is a tensor with only 1 entry,
+        # token_ids is a torch.Tensor
+        token_ids = words[word_ids]
+        if isinstance(token_ids, k2.RaggedTensor):
+            token_ids = token_ids.values
+
+        # token_ids is a 1-D tensor containing the BPE tokens
+        # of the current sentence
+
+        sentence_lengths[i] = token_ids.numel()
+
+    output["sentence_lengths"] = torch.tensor(sentence_lengths, dtype=torch.int32)
+
+    torch.save(output, args.lm_archive)
+    logging.info(f"Saved to {args.lm_archive}")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/swbd/ASR/local/prepare_swbd_testsets.py

@@ -0,0 +1,161 @@
+import argparse
+import logging
+import os
+from pathlib import Path
+import tarfile
+from itertools import chain
+from typing import Dict, Optional, Union
+
+from lhotse import fix_manifests, validate_recordings_and_supervisions
+from lhotse.audio import Recording, RecordingSet
+from lhotse.supervision import SupervisionSegment, SupervisionSet
+from lhotse.utils import Pathlike, check_and_rglob, resumable_download, safe_extract
+
+import sentencepiece as spm
+from filter_cuts import filter_cuts
+from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
+from lhotse.recipes.utils import read_manifests_if_cached
+
+from icefall.utils import get_executor, str2bool
+
+
+def prepare_switchboard(
+    audio_dir: Pathlike,
+    transcripts_dir: Optional[Pathlike] = None,
+    sentiment_dir: Optional[Pathlike] = None,
+    output_dir: Optional[Pathlike] = None,
+    omit_silence: bool = True,
+    absolute_paths: bool = False,
+) -> Dict[str, Union[RecordingSet, SupervisionSet]]:
+    """
+    Prepare manifests for the Switchboard corpus.
+    We create two manifests: one with recordings, and the other one with text supervisions.
+    When ``sentiment_dir`` is provided, we create another supervision manifest with sentiment annotations.
+    :param audio_dir: Path to ``LDC97S62`` package.
+    :param transcripts_dir: Path to the transcripts directory (typically named "swb_ms98_transcriptions").
+        If not provided, the transcripts will be downloaded.
+    :param sentiment_dir: Optional path to ``LDC2020T14`` package which contains sentiment annotations
+        for SWBD segments.
+    :param output_dir: Directory where the manifests should be written. Can be omitted to avoid writing.
+    :param omit_silence: Whether supervision segments with ``[silence]`` token should be removed or kept.
+    :param absolute_paths: Whether to return absolute or relative (to the corpus dir) paths for recordings.
+    :return: A dict with manifests. The keys are: ``{'recordings', 'supervisions'}``.
+    """
+    audio_paths = check_and_rglob(audio_dir, "*.sph")
+    text_paths = check_and_rglob(transcripts_dir, "*trans.text")
+
+    groups = []
+    name_to_text = {p.stem.split("-")[0]: p for p in text_paths}
+    for ap in audio_paths:
+        name = ap.stem.replace("sw0", "sw")
+        groups.append(
+            {
+                "audio": ap,
+                "text-0": name_to_text[f"{name}A"],
+                "text-1": name_to_text[f"{name}B"],
+            }
+        )
+
+    recordings = RecordingSet.from_recordings(
+        Recording.from_file(
+            group["audio"], relative_path_depth=None if absolute_paths else 3
+        )
+        for group in groups
+    )
+    supervisions = SupervisionSet.from_segments(
+        chain.from_iterable(
+            make_segments(
+                transcript_path=group[f"text-{channel}"],
+                recording=recording,
+                channel=channel,
+                omit_silence=omit_silence,
+            )
+            for group, recording in zip(groups, recordings)
+            for channel in [0, 1]
+        )
+    )
+
+    recordings, supervisions = fix_manifests(recordings, supervisions)
+    validate_recordings_and_supervisions(recordings, supervisions)
+
+    if sentiment_dir is not None:
+        parse_and_add_sentiment_labels(sentiment_dir, supervisions)
+
+    if output_dir is not None:
+        output_dir = Path(output_dir)
+        output_dir.mkdir(parents=True, exist_ok=True)
+        recordings.to_file(output_dir / "swbd_recordings_all.jsonl.gz")
+        supervisions.to_file(output_dir / "swbd_supervisions_all.jsonl.gz")
+    return {"recordings": recordings, "supervisions": supervisions}
+
+
+def make_segments(
+    transcript_path: Path, recording: Recording, channel: int, omit_silence: bool = True
+):
+    lines = transcript_path.read_text().splitlines()
+    return [
+        SupervisionSegment(
+            id=segment_id,
+            recording_id=recording.id,
+            start=float(start),
+            duration=round(float(end) - float(start), ndigits=8),
+            channel=channel,
+            text=" ".join(words),
+            language="English",
+            speaker=f"{recording.id}A",
+        )
+        for segment_id, start, end, *words in map(str.split, lines)
+        if words[0] != "[silence]" or not omit_silence
+    ]
+
+
+def download_and_untar(
+    target_dir: Pathlike = ".", force_download: bool = False, url: str = SWBD_TEXT_URL
+) -> Path:
+    target_dir = Path(target_dir)
+    transcript_dir = target_dir / "swb_ms98_transcriptions"
+    if transcript_dir.is_dir():
+        return transcript_dir
+    target_dir.mkdir(parents=True, exist_ok=True)
+    tar_name = "switchboard_word_alignments.tar.gz"
+    tar_path = target_dir / tar_name
+    resumable_download(url, filename=tar_path, force_download=force_download)
+    with tarfile.open(tar_path) as tar:
+        safe_extract(tar, path=target_dir)
+    return transcript_dir
+
+
+def parse_and_add_sentiment_labels(
+    sentiment_dir: Pathlike, supervisions: SupervisionSet
+):
+    """Read 'LDC2020T14' sentiment annotations and add then to the supervision segments."""
+    import pandas as pd
+
+    # Sanity checks
+    sentiment_dir = Path(sentiment_dir)
+    labels = sentiment_dir / "data" / "sentiment_labels.tsv"
+    assert sentiment_dir.is_dir() and labels.is_file()
+    # Read the TSV as a dataframe
+    df = pd.read_csv(labels, delimiter="\t", names=["id", "start", "end", "sentiment"])
+    # We are going to match the segments in LDC2020T14 with the ones we already
+    # parsed from ISIP transcripts. We simply look which of the existing segments
+    # fall into a sentiment-annotated time span. When doing it this way, we use
+    # 51773 out of 52293 available sentiment annotations, which should be okay.
+    for _, row in df.iterrows():
+        call_id = row["id"].split("_")[0]
+        matches = list(
+            supervisions.find(
+                recording_id=call_id,
+                start_after=row["start"] - 1e-2,
+                end_before=row["end"] + 1e-2,
+            )
+        )
+        if not matches:
+            continue
+        labels = row["sentiment"].split("#")
+        # SupervisionSegments returned from .find() are references to the ones in the
+        # SupervisionSet, so we can just modify them. We use the "custom" field
+        # to add the sentiment label. Since there were multiple annotators,
+        # we add all available labels and leave it up to the user to disambiguate them.
+        for segment in matches:
+            segment.custom = {f"sentiment{i}": label for i, label in enumerate(labels)}

egs/swbd/ASR/local/rt03_data_prep.sh

@@ -0,0 +1,107 @@
+#!/usr/bin/env bash
+
+# RT-03 data preparation (conversational telephone speech part only)
+# Adapted from Arnab Ghoshal's script for Hub-5 Eval 2000 by Peng Qi
+
+# To be run from one directory above this script.
+
+# Expects the standard directory layout for RT-03
+
+if [ $# -ne 1 ]; then
+    echo "Usage: $0 <rt03-dir>"
+    echo "e.g.: $0 /export/corpora/LDC/LDC2007S10"
+    echo "See comments in the script for more details"
+    exit 1
+fi
+
+sdir=$1
+[ ! -d $sdir/data/audio/eval03/english/cts ] \
+    && echo Expecting directory $sdir/data/audio/eval03/english/cts to be present && exit 1;
+    [ ! -d $sdir/data/references/eval03/english/cts ] \
+        && echo Expecting directory $tdir/data/references/eval03/english/cts to be present && exit 1;
+
+    dir=data/local/rt03
+    mkdir -p $dir
+
+    rtroot=$sdir
+    tdir=$sdir/data/references/eval03/english/cts
+    sdir=$sdir/data/audio/eval03/english/cts
+
+    find -L $sdir -iname '*.sph' | sort > $dir/sph.flist
+    sed -e 's?.*/??' -e 's?.sph??' $dir/sph.flist | paste - $dir/sph.flist \
+        > $dir/sph.scp
+
+    sph2pipe=sph2pipe
+    ! command -v "${sph2pipe}" &> /dev/null \
+        && echo "Could not execute the sph2pipe program at $sph2pipe" && exit 1;
+
+    awk -v sph2pipe=$sph2pipe '{
+    printf("%s-A %s -f wav -p -c 1 %s |\n", $1, sph2pipe, $2);
+    printf("%s-B %s -f wav -p -c 2 %s |\n", $1, sph2pipe, $2);
+}' < $dir/sph.scp | sort > $dir/wav.scp || exit 1;
+#side A - channel 1, side B - channel 2
+
+# Get segments file...
+# segments file format is: utt-id side-id start-time end-time, e.g.:
+# sw02001-A_000098-001156 sw02001-A 0.98 11.56
+#pem=$sdir/english/hub5e_00.pem
+#[ ! -f $pem ] && echo "No such file $pem" && exit 1;
+# pem file has lines like:
+# en_4156 A unknown_speaker 301.85 302.48
+
+#grep -v ';;' $pem \
+cat $tdir/*.stm | grep -v ';;' | grep -v inter_segment_gap \
+    | awk '{
+    spk=$1"-"(($2==1)?"A":"B");
+    utt=sprintf("%s_%06d-%06d",spk,$4*100,$5*100);
+    print utt,spk,$4,$5;}' \
+        | sort -u > $dir/segments
+
+# stm file has lines like:
+# en_4156 A en_4156_A 357.64 359.64 <O,en,F,en-F>  HE IS A POLICE OFFICER
+# TODO(arnab): We should really be lowercasing this since the Edinburgh
+# recipe uses lowercase. This is not used in the actual scoring.
+#grep -v ';;' $tdir/reference/hub5e00.english.000405.stm \
+cat $tdir/*.stm | grep -v ';;' | grep -v inter_segment_gap \
+    | awk '{
+    spk=$1"-"(($2==1)?"A":"B");
+    utt=sprintf("%s_%06d-%06d",spk,$4*100,$5*100);
+    printf utt; for(n=7;n<=NF;n++) printf(" %s", $n); print ""; }' \
+        | sort > $dir/text.all
+
+# We'll use the stm file for sclite scoring.  There seem to be various errors
+# in the stm file that upset hubscr.pl, and we fix them here.
+cat $tdir/*.stm | \
+    sed -e 's:((:(:' -e 's:<B_ASIDE>::g' -e 's:<E_ASIDE>::g' | \
+    grep -v inter_segment_gap | \
+    awk '{
+    printf $1; if ($1==";;") printf(" %s",$2); else printf(($2==1)?" A":" B"); for(n=3;n<=NF;n++) printf(" %s", $n); print ""; }'\
+        > $dir/stm
+            #$tdir/reference/hub5e00.english.000405.stm >  $dir/stm
+            cp $rtroot/data/trans_rules/en20030506.glm  $dir/glm
+
+# next line uses command substitution
+# Just checking that the segments are the same in pem vs. stm.
+! cmp <(awk '{print $1}' $dir/text.all) <(awk '{print $1}' $dir/segments) && \
+    echo "Segments from pem file and stm file do not match." && exit 1;
+
+grep -v IGNORE_TIME_SEGMENT_ $dir/text.all > $dir/text
+
+# create an utt2spk file that assumes each conversation side is
+# a separate speaker.
+awk '{print $1,$2;}' $dir/segments > $dir/utt2spk
+utils/utt2spk_to_spk2utt.pl $dir/utt2spk > $dir/spk2utt
+
+# cp $dir/segments $dir/segments.tmp
+# awk '{x=$3-0.05; if (x<0.0) x=0.0; y=$4+0.05; print $1, $2, x, y; }' \
+#   $dir/segments.tmp > $dir/segments
+
+awk '{print $1}' $dir/wav.scp \
+    | perl -ane '$_ =~ m:^(\S+)-([AB])$: || die "bad label $_";
+    print "$1-$2 $1 $2\n"; ' \
+        > $dir/reco2file_and_channel || exit 1;
+
+    ./utils/fix_data_dir.sh $dir
+
+    echo Data preparation and formatting completed for RT-03
+    echo "(but not MFCC extraction)"

egs/swbd/ASR/local/sort_lm_training_data.py

@@ -0,0 +1,141 @@
+#!/usr/bin/env python3
+
+# Copyright (c)  2021  Xiaomi Corporation (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+This file takes as input the filename of LM training data
+generated by ./local/prepare_lm_training_data.py and sorts
+it by sentence length.
+
+Sentence length equals to the number of BPE tokens in a sentence.
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import k2
+import numpy as np
+import torch
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--in-lm-data",
+        type=str,
+        help="Input LM training data, e.g., data/bpe_500/lm_data.pt",
+    )
+
+    parser.add_argument(
+        "--out-lm-data",
+        type=str,
+        help="Input LM training data, e.g., data/bpe_500/sorted_lm_data.pt",
+    )
+
+    parser.add_argument(
+        "--out-statistics",
+        type=str,
+        help="Statistics about LM training data., data/bpe_500/statistics.txt",
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    in_lm_data = Path(args.in_lm_data)
+    out_lm_data = Path(args.out_lm_data)
+    assert in_lm_data.is_file(), f"{in_lm_data}"
+    if out_lm_data.is_file():
+        logging.warning(f"{out_lm_data} exists - skipping")
+        return
+    data = torch.load(in_lm_data)
+    words2bpe = data["words"]
+    sentences = data["sentences"]
+    sentence_lengths = data["sentence_lengths"]
+
+    num_sentences = sentences.dim0
+    assert num_sentences == sentence_lengths.numel(), (
+        num_sentences,
+        sentence_lengths.numel(),
+    )
+
+    indices = torch.argsort(sentence_lengths, descending=True)
+
+    sorted_sentences = sentences[indices.to(torch.int32)]
+    sorted_sentence_lengths = sentence_lengths[indices]
+
+    # Check that sentences are ordered by length
+    assert num_sentences == sorted_sentences.dim0, (
+        num_sentences,
+        sorted_sentences.dim0,
+    )
+
+    cur = None
+    for i in range(num_sentences):
+        word_ids = sorted_sentences[i]
+        token_ids = words2bpe[word_ids]
+        if isinstance(token_ids, k2.RaggedTensor):
+            token_ids = token_ids.values
+        if cur is not None:
+            assert cur >= token_ids.numel(), (cur, token_ids.numel())
+
+        cur = token_ids.numel()
+        assert cur == sorted_sentence_lengths[i]
+
+    data["sentences"] = sorted_sentences
+    data["sentence_lengths"] = sorted_sentence_lengths
+    torch.save(data, args.out_lm_data)
+    logging.info(f"Saved to {args.out_lm_data}")
+
+    statistics = Path(args.out_statistics)
+
+    # Write statistics
+    num_words = sorted_sentences.numel()
+    num_tokens = sentence_lengths.sum().item()
+    max_sentence_length = sentence_lengths[indices[0]]
+    min_sentence_length = sentence_lengths[indices[-1]]
+
+    step = 10
+    hist, bins = np.histogram(
+        sentence_lengths.numpy(),
+        bins=np.arange(1, max_sentence_length + step, step),
+    )
+
+    histogram = np.stack((bins[:-1], hist)).transpose()
+
+    with open(statistics, "w") as f:
+        f.write(f"num_sentences: {num_sentences}\n")
+        f.write(f"num_words: {num_words}\n")
+        f.write(f"num_tokens: {num_tokens}\n")
+        f.write(f"max_sentence_length: {max_sentence_length}\n")
+        f.write(f"min_sentence_length: {min_sentence_length}\n")
+        f.write("histogram:\n")
+        f.write("  bin  count  percent\n")
+        for row in histogram:
+            f.write(
+                f"{int(row[0]):>5} {int(row[1]):>5}   "
+                f"{100.*row[1]/num_sentences:.3f}%\n"
+            )
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/swbd/ASR/local/swbd1_data_prep.sh

@@ -0,0 +1,132 @@
+#!/usr/bin/env bash
+
+# Switchboard-1 training data preparation customized for Edinburgh
+# Author:  Arnab Ghoshal (Jan 2013)
+
+# To be run from one directory above this script.
+
+## The input is some directory containing the switchboard-1 release 2
+## corpus (LDC97S62).  Note: we don't make many assumptions about how
+## you unpacked this.  We are just doing a "find" command to locate
+## the .sph files.
+
+## The second input is optional, which should point to a directory containing
+## Switchboard transcriptions/documentations (specifically, the conv.tab file).
+## If specified, the script will try to use the actual speaker PINs provided
+## with the corpus instead of the conversation side ID (Kaldi default). We
+## will be using "find" to locate this file so we don't make any assumptions
+## on the directory structure. (Peng Qi, Aug 2014)
+
+
+#check existing directories
+if [ $# != 1 -a $# != 2 ]; then
+    echo "Usage: swbd1_data_prep.sh /path/to/SWBD [/path/to/SWBD_DOC]"
+    exit 1;
+fi
+
+SWBD_DIR=$1
+
+dir=data/local/train
+mkdir -p $dir
+
+
+# Audio data directory check
+if [ ! -d $SWBD_DIR ]; then
+    echo "Error: run.sh requires a directory argument"
+    exit 1;
+fi
+
+sph2pipe=sph2pipe
+! command -v "${sph2pipe}" &> /dev/null \
+    && echo "Could not execute the sph2pipe program at $sph2pipe" && exit 1;
+
+# Option A: SWBD dictionary file check
+[ ! -f ./swb_ms98_transcriptions/sw-ms98-dict.text ] && \
+    echo  "SWBD dictionary file does not exist" &&  exit 1;
+
+# find sph audio files
+find -L $SWBD_DIR -iname '*.sph' | sort > $dir/sph.flist
+
+n=`cat $dir/sph.flist | wc -l`
+[ $n -ne 2435 ] && [ $n -ne 2438 ] && \
+    echo Warning: expected 2435 or 2438 data data files, found $n
+
+
+# (1a) Transcriptions preparation
+# make basic transcription file (add segments info)
+# **NOTE: In the default Kaldi recipe, everything is made uppercase, while we
+# make everything lowercase here. This is because we will be using SRILM which
+# can optionally make everything lowercase (but not uppercase) when mapping
+# LM vocabs.
+awk '{
+name=substr($1,1,6); gsub("^sw","sw0",name); side=substr($1,7,1);
+stime=$2; etime=$3;
+printf("%s-%s_%06.0f-%06.0f",
+name, side, int(100*stime+0.5), int(100*etime+0.5));
+for(i=4;i<=NF;i++) printf(" %s", $i); printf "\n"
+}' ./swb_ms98_transcriptions/*/*/*-trans.text  > $dir/transcripts1.txt
+
+# test if trans. file is sorted
+export LC_ALL=C;
+sort -c $dir/transcripts1.txt || exit 1; # check it's sorted.
+
+# Remove SILENCE, <B_ASIDE> and <E_ASIDE>.
+
+# Note: we have [NOISE], [VOCALIZED-NOISE], [LAUGHTER], [SILENCE].
+# removing [SILENCE], and the <B_ASIDE> and <E_ASIDE> markers that mark
+# speech to somone; we will give phones to the other three (NSN, SPN, LAU).
+# There will also be a silence phone, SIL.
+# **NOTE: modified the pattern matches to make them case insensitive
+cat $dir/transcripts1.txt \
+    | perl -ane 's:\s\[SILENCE\](\s|$):$1:gi;
+    s/<B_ASIDE>//gi;
+    s/<E_ASIDE>//gi;
+    print;' \
+        | awk '{if(NF > 1) { print; } } ' > $dir/transcripts2.txt
+
+
+# **NOTE: swbd1_map_words.pl has been modified to make the pattern matches
+# case insensitive
+local/swbd1_map_words.pl -f 2- $dir/transcripts2.txt  > $dir/text
+
+# format acronyms in text
+python3 local/map_acronyms_transcripts.py -i $dir/text -o $dir/text_map \
+    -M data/local/dict_nosp/acronyms.map
+    mv $dir/text_map $dir/text
+
+# (1c) Make segment files from transcript
+#segments file format is: utt-id side-id start-time end-time, e.g.:
+#sw02001-A_000098-001156 sw02001-A 0.98 11.56
+awk '{
+segment=$1;
+split(segment,S,"[_-]");
+side=S[2]; audioname=S[1]; startf=S[3]; endf=S[4];
+print segment " " audioname "-" side " " startf/100 " " endf/100
+}' < $dir/text > $dir/segments
+
+sed -e 's?.*/??' -e 's?.sph??' $dir/sph.flist | paste - $dir/sph.flist \
+    > $dir/sph.scp
+
+awk -v sph2pipe=$sph2pipe '{
+printf("%s-A %s -f wav -p -c 1 %s |\n", $1, sph2pipe, $2);
+printf("%s-B %s -f wav -p -c 2 %s |\n", $1, sph2pipe, $2);
+}' < $dir/sph.scp | sort > $dir/wav.scp || exit 1;
+#side A - channel 1, side B - channel 2
+
+# this file reco2file_and_channel maps recording-id (e.g. sw02001-A)
+# to the file name sw02001 and the A, e.g.
+# sw02001-A  sw02001 A
+# In this case it's trivial, but in other corpora the information might
+# be less obvious.  Later it will be needed for ctm scoring.
+awk '{print $1}' $dir/wav.scp \
+    | perl -ane '$_ =~ m:^(\S+)-([AB])$: || die "bad label $_";
+    print "$1-$2 $1 $2\n"; ' \
+        > $dir/reco2file_and_channel || exit 1;
+
+    awk '{spk=substr($1,1,9); print $1 " " spk}' $dir/segments > $dir/utt2spk \
+        || exit 1;
+            sort -k 2 $dir/utt2spk | utils/utt2spk_to_spk2utt.pl > $dir/spk2utt || exit 1;
+
+            echo Switchboard-1 data preparation succeeded.
+
+            utils/fix_data_dir.sh data/local/train

egs/swbd/ASR/local/swbd1_map_words.pl

@@ -0,0 +1,52 @@
+#!/usr/bin/env perl
+
+# Modified from swbd_map_words.pl in Kaldi s5 recipe to make pattern
+# matches case-insensitive --Arnab (Jan 2013)
+
+if ($ARGV[0] eq "-f") {
+  shift @ARGV;
+  $field_spec = shift @ARGV;
+  if ($field_spec =~ m/^\d+$/) {
+    $field_begin = $field_spec - 1; $field_end = $field_spec - 1;
+  }
+  if ($field_spec =~ m/^(\d*)[-:](\d*)/) { # accept e.g. 1:10 as a courtesy (properly, 1-10)
+    if ($1 ne "") {
+      $field_begin = $1 - 1;    # Change to zero-based indexing.
+    }
+    if ($2 ne "") {
+      $field_end = $2 - 1;      # Change to zero-based indexing.
+    }
+  }
+  if (!defined $field_begin && !defined $field_end) {
+    die "Bad argument to -f option: $field_spec";
+  }
+}
+
+
+while (<>) {
+  @A = split(" ", $_);
+  for ($n = 0; $n < @A; $n++) {
+    $a = $A[$n];
+    if ( (!defined $field_begin || $n >= $field_begin)
+         && (!defined $field_end || $n <= $field_end)) {
+      # e.g. [LAUGHTER-STORY] -> STORY;
+      $a =~ s:(|\-)^\[LAUGHTER-(.+)\](|\-)$:$1$2$3:i;
+      # $1 and $3 relate to preserving trailing "-"
+      $a =~ s:^\[(.+)/.+\](|\-)$:$1$2:; # e.g. [IT'N/ISN'T] -> IT'N ... note,
+      # 1st part may include partial-word stuff, which we process further below,
+      # e.g. [LEM[GUINI]-/LINGUINI]
+      # the (|\_) at the end is to accept and preserve trailing -'s.
+      $a =~ s:^(|\-)\[[^][]+\](.+)$:-$2:; # e.g. -[AN]Y , note \047 is quote;
+      # let the leading - be optional on input, as sometimes omitted.
+      $a =~ s:^(.+)\[[^][]+\](|\-)$:$1-:; # e.g. AB[SOLUTE]- -> AB-;
+      # let the trailing - be optional on input, as sometimes omitted.
+      $a =~ s:([^][]+)\[.+\]$:$1:; # e.g. EX[SPECIALLY]-/ESPECIALLY] -> EX-
+      # which is a  mistake in the input.
+      $a =~ s:^\{(.+)\}$:$1:;                 # e.g. {YUPPIEDOM} -> YUPPIEDOM
+      $a =~ s:[A-Z]\[([^][])+\][A-Z]:$1-$3:i; # e.g. AMMU[N]IT- -> AMMU-IT-
+      $a =~ s:_\d$::;                         # e.g. THEM_1 -> THEM
+    }
+    $A[$n] = $a;
+  }
+  print join(" ", @A) . "\n";
+}

egs/swbd/ASR/local/swbd1_prepare_dict.sh

@@ -0,0 +1,92 @@
+#!/usr/bin/env bash
+
+# Formatting the Mississippi State dictionary for use in Edinburgh. Differs
+# from the one in Kaldi s5 recipe in that it uses lower-case --Arnab (Jan 2013)
+
+# To be run from one directory above this script.
+
+
+#check existing directories
+[ $# != 0 ] && echo "Usage: local/swbd1_data_prep.sh" && exit 1;
+
+srcdir=.  # This is where we downloaded some stuff..
+dir=./data/local/dict_nosp
+mkdir -p $dir
+srcdict=$srcdir/swb_ms98_transcriptions/sw-ms98-dict.text
+
+# assume swbd_p1_data_prep.sh was done already.
+[ ! -f "$srcdict" ] && echo "$0: No such file $srcdict" && exit 1;
+
+cp $srcdict $dir/lexicon0.txt || exit 1;
+chmod a+w $dir/lexicon0.txt
+patch <local/dict.patch $dir/lexicon0.txt || exit 1;
+
+#(2a) Dictionary preparation:
+# Pre-processing (remove comments)
+grep -v '^#' $dir/lexicon0.txt | awk 'NF>0' | sort > $dir/lexicon1.txt || exit 1;
+
+cat $dir/lexicon1.txt | awk '{ for(n=2;n<=NF;n++){ phones[$n] = 1; }} END{for (p in phones) print p;}' | \
+    grep -v sil > $dir/nonsilence_phones.txt  || exit 1;
+
+( echo sil; echo spn; echo nsn; echo lau ) > $dir/silence_phones.txt
+
+echo sil > $dir/optional_silence.txt
+
+# No "extra questions" in the input to this setup, as we don't
+# have stress or tone.
+echo -n >$dir/extra_questions.txt
+
+cp local/MSU_single_letter.txt $dir/
+# Add to the lexicon the silences, noises etc.
+# Add single letter lexicon
+# The original swbd lexicon does not have precise single letter lexicion
+# e.g. it does not have entry of W
+( echo '!sil sil'; echo '[vocalized-noise] spn'; echo '[noise] nsn'; \
+    echo '[laughter] lau'; echo '<unk> spn' ) \
+    | cat - $dir/lexicon1.txt $dir/MSU_single_letter.txt  > $dir/lexicon2.txt || exit 1;
+
+# Map the words in the lexicon.  That is-- for each word in the lexicon, we map it
+# to a new written form.  The transformations we do are:
+# remove laughter markings, e.g.
+# [LAUGHTER-STORY] -> STORY
+# Remove partial-words, e.g.
+# -[40]1K W AH N K EY
+# becomes -1K
+# and
+# -[AN]Y IY
+# becomes
+# -Y
+# -[A]B[OUT]- B
+# becomes
+# -B-
+# Also, curly braces, which appear to be used for "nonstandard"
+# words or non-words, are removed, e.g.
+# {WOLMANIZED} W OW L M AX N AY Z D
+# -> WOLMANIZED
+# Also, mispronounced words, e.g.
+#  [YEAM/YEAH] Y AE M
+# are changed to just e.g. YEAM, i.e. the orthography
+# of the mispronounced version.
+# Note-- this is only really to be used in training.  The main practical
+# reason is to avoid having tons of disambiguation symbols, which
+# we otherwise would get because there are many partial words with
+# the same phone sequences (most problematic: S).
+# Also, map
+# THEM_1 EH M -> THEM
+# so that multiple pronunciations just have alternate entries
+# in the lexicon.
+
+local/swbd1_map_words.pl -f 1 $dir/lexicon2.txt | sort -u \
+    > $dir/lexicon3.txt || exit 1;
+
+python3 local/format_acronyms_dict.py -i $dir/lexicon3.txt -o $dir/lexicon4.txt \
+    -L $dir/MSU_single_letter.txt -M $dir/acronyms_raw.map
+    cat $dir/acronyms_raw.map | sort -u > $dir/acronyms.map
+
+    ( echo 'i ay' )| cat - $dir/lexicon4.txt | tr '[A-Z]' '[a-z]' | sort -u > $dir/lexicon5.txt
+
+    pushd $dir >&/dev/null
+    ln -sf lexicon5.txt lexicon.txt # This is the final lexicon.
+    popd >&/dev/null
+    rm $dir/lexiconp.txt 2>/dev/null
+    echo Prepared input dictionary and phone-sets for Switchboard phase 1.

egs/swbd/ASR/local/train_bpe_model.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+# You can install sentencepiece via:
+#
+#  pip install sentencepiece
+#
+# Due to an issue reported in
+# https://github.com/google/sentencepiece/pull/642#issuecomment-857972030
+#
+# Please install a version >=0.1.96
+
+import argparse
+import shutil
+from pathlib import Path
+
+import sentencepiece as spm
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        help="""Input and output directory.
+        The generated bpe.model is saved to this directory.
+        """,
+    )
+
+    parser.add_argument(
+        "--transcript",
+        type=str,
+        help="Training transcript.",
+    )
+
+    parser.add_argument(
+        "--vocab-size",
+        type=int,
+        help="Vocabulary size for BPE training",
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    vocab_size = args.vocab_size
+    lang_dir = Path(args.lang_dir)
+
+    model_type = "unigram"
+
+    model_prefix = f"{lang_dir}/{model_type}_{vocab_size}"
+    train_text = args.transcript
+    character_coverage = 1.0
+    input_sentence_size = 100000000
+
+    user_defined_symbols = ["<blk>", "<sos/eos>"]
+    unk_id = len(user_defined_symbols)
+    # Note: unk_id is fixed to 2.
+    # If you change it, you should also change other
+    # places that are using it.
+
+    model_file = Path(model_prefix + ".model")
+    if not model_file.is_file():
+        spm.SentencePieceTrainer.train(
+            input=train_text,
+            vocab_size=vocab_size,
+            model_type=model_type,
+            model_prefix=model_prefix,
+            input_sentence_size=input_sentence_size,
+            character_coverage=character_coverage,
+            user_defined_symbols=user_defined_symbols,
+            unk_id=unk_id,
+            bos_id=-1,
+            eos_id=-1,
+        )
+    else:
+        print(f"{model_file} exists - skipping")
+        return
+
+    shutil.copyfile(model_file, f"{lang_dir}/bpe.model")
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/local/validate_bpe_lexicon.py

@@ -0,0 +1,77 @@
+#!/usr/bin/env python3
+# Copyright    2022  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+This script checks that there are no OOV tokens in the BPE-based lexicon.
+
+Usage example:
+
+    python3 ./local/validate_bpe_lexicon.py \
+            --lexicon /path/to/lexicon.txt \
+            --bpe-model /path/to/bpe.model
+"""
+
+import argparse
+from pathlib import Path
+from typing import List, Tuple
+
+import sentencepiece as spm
+
+from icefall.lexicon import read_lexicon
+
+# Map word to word pieces
+Lexicon = List[Tuple[str, List[str]]]
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--lexicon",
+        required=True,
+        type=Path,
+        help="Path to lexicon.txt",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        required=True,
+        type=Path,
+        help="Path to bpe.model",
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    assert args.lexicon.is_file(), args.lexicon
+    assert args.bpe_model.is_file(), args.bpe_model
+
+    lexicon = read_lexicon(args.lexicon)
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(str(args.bpe_model))
+
+    word_pieces = set(sp.id_to_piece(list(range(sp.vocab_size()))))
+    for word, pieces in lexicon:
+        for p in pieces:
+            if p not in word_pieces:
+                raise ValueError(f"The word {word} contains an OOV token {p}")
+
+
+if __name__ == "__main__":
+    main()

egs/swbd/ASR/prepare.sh

@@ -0,0 +1,403 @@
+#!/usr/bin/env bash
+
+# fix segmentation fault reported in https://github.com/k2-fsa/icefall/issues/674
+export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
+
+set -eou pipefail
+
+nj=15
+stage=-1
+stop_stage=100
+
+# We assume dl_dir (download dir) contains the following
+# directories and files. If not, they will be downloaded
+# by this script automatically.
+#
+#  - $dl_dir/musan
+#      This directory contains the following directories downloaded from
+#       http://www.openslr.org/17/
+#
+#     - music
+#     - noise
+#     - speech
+
+dl_dir=./download
+swbd1_dir="/export/corpora3/LDC/LDC97S62"
+eval2000_dir="/export/corpora2/LDC/LDC2002S09/hub5e_00"
+eval2000_ref_dir="/export/corpora2/LDC/LDC2002T43"
+rt03_dir="/export/corpora/LDC/LDC2007S10"
+
+. shared/parse_options.sh || exit 1
+
+# vocab size for sentence piece models.
+# It will generate data/lang_bpe_xxx,
+# data/lang_bpe_yyy if the array contains xxx, yyy
+vocab_sizes=(
+    5000
+    2000
+    1000
+    500
+)
+
+# All files generated by this script are saved in "data".
+# You can safely remove "data" and rerun this script to regenerate it.
+mkdir -p data
+
+log() {
+    # This function is from espnet
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+log "swbd1_dir: $swbd1_dir"
+log "eval2000_dir: $eval2000_dir $eval2000_ref_dir"
+log "rt03_dir: $rt03_dir"
+
+if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
+    log "Stage 1: Prepare SwitchBoard manifest"
+    # We assume that you have downloaded the SwitchBoard corpus
+    # to respective dirs
+    mkdir -p data/manifests
+    if [ ! -e data/manifests/.swbd.done ]; then
+        lhotse prepare switchboard --absolute-paths True $swbd1_dir data/manifests_train
+        ./local/swbd1_prepare_dict.sh
+        ./local/swbd1_data_prep.sh $swbd1_dir
+        lhotse kaldi import data/local/train 8000 data/manifests_train
+        mv data/manifests_train/recordings.jsonl.gz data/manifests_train/swbd_recordings_all.jsonl.gz
+        mv data/manifests_train/supervisions.jsonl.gz data/manifests_train/swbd_supervisions_all.jsonl.gz
+
+        ./local/eval2000_data_prep.sh $eval2000_dir $eval2000_ref_dir
+        ./local/rt03_data_prep.sh $rt03_dir
+
+        # normalize eval2000 and rt03 texts by
+        # 1) convert upper to lower
+        # 2) remove tags (%AH) (%HESITATION) (%UH)
+        # 3) remove <B_ASIDE> <E_ASIDE>
+        # 4) remove "(" or ")"
+        for x in eval2000 rt03; do
+            cp data/local/${x}/text data/local/${x}/text.org
+            paste -d "" \
+                <(cut -f 1 -d" " data/local/${x}/text.org) \
+                <(awk '{$1=""; print tolower($0)}' data/local/${x}/text.org | perl -pe 's| \(\%.*\)||g' | perl -pe 's| \<.*\>||g' | sed -e "s/(//g" -e "s/)//g") \
+                | sed -e 's/\s\+/ /g' > data/local/${x}/text
+                            rm data/local/${x}/text.org
+                        done
+
+                        python ./local/filter_empty_text.py --kaldi-data-dir data/local/eval2000
+                        ./utils/fix_data_dir.sh data/local/eval2000
+                        lhotse kaldi import data/local/eval2000 8000 data/manifests_eval2000
+                        mv data/manifests_eval2000/recordings.jsonl.gz data/manifests_eval2000/swbd_recordings_eval2000.jsonl.gz
+                        mv data/manifests_eval2000/supervisions.jsonl.gz data/manifests_eval2000/swbd_supervisions_eval2000.jsonl.gz
+
+                        python ./local/filter_empty_text.py --kaldi-data-dir data/local/rt03
+                        ./utils/fix_data_dir.sh data/local/rt03
+                        lhotse kaldi import data/local/rt03 8000 data/manifests_rt03
+                        mv data/manifests_rt03/recordings.jsonl.gz data/manifests_rt03/swbd_recordings_rt03.jsonl.gz
+                        mv data/manifests_rt03/supervisions.jsonl.gz data/manifests_rt03/swbd_supervisions_rt03.jsonl.gz
+
+                        lhotse fix data/manifests_train/swbd_recordings_all.jsonl.gz data/manifests_train/swbd_supervisions_all.jsonl.gz data/manifests
+                        lhotse fix data/manifests_eval2000/swbd_recordings_eval2000.jsonl.gz data/manifests_eval2000/swbd_supervisions_eval2000.jsonl.gz data/manifests
+                        lhotse fix data/manifests_rt03/swbd_recordings_rt03.jsonl.gz data/manifests_rt03/swbd_supervisions_rt03.jsonl.gz data/manifests
+
+                        touch data/manifests/.swbd.done
+    fi
+fi
+
+if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
+    log "Stage 2: Prepare musan manifest"
+    # We assume that you have downloaded the musan corpus
+    # to $dl_dir/musan
+    mkdir -p data/manifests
+    if [ ! -e data/manifests/.musan.done ]; then
+        lhotse prepare musan $dl_dir/musan data/manifests
+        touch data/manifests/.musan.done
+    fi
+fi
+
+if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
+    log "Stage 3: Compute fbank for switchboard"
+    mkdir -p data/fbank
+    if [ ! -e data/fbank/.swbd.done ]; then
+        ./local/compute_fbank_swbd.py
+        touch data/fbank/.swbd.done
+    fi
+fi
+
+if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
+    log "Stage 4: Compute fbank for musan"
+    mkdir -p data/fbank
+    if [ ! -e data/fbank/.musan.done ]; then
+        ./local/compute_fbank_musan.py
+        touch data/fbank/.musan.done
+    fi
+fi
+
+if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
+    log "Stage 5: Prepare phone based lang"
+    lang_dir=data/lang_phone
+    mkdir -p $lang_dir
+
+    log "prepare dict"
+    cut -f 2- -d" " data/local/train/text >${lang_dir}/input.txt
+    # [noise] nsn
+    # !sil sil
+    # <unk> spn
+    cat data/local/dict_nosp/lexicon.txt |
+        sort | uniq >$lang_dir/lexicon.txt
+
+    if [ ! -f $lang_dir/L_disambig.pt ]; then
+        ./local/prepare_lang.py --lang-dir $lang_dir
+    fi
+
+    if [ ! -f $lang_dir/L.fst ]; then
+        log "Converting L.pt to L.fst"
+        ./shared/convert-k2-to-openfst.py \
+            --olabels aux_labels \
+            $lang_dir/L.pt \
+            $lang_dir/L.fst
+    fi
+
+    if [ ! -f $lang_dir/L_disambig.fst ]; then
+        log "Converting L_disambig.pt to L_disambig.fst"
+        ./shared/convert-k2-to-openfst.py \
+            --olabels aux_labels \
+            $lang_dir/L_disambig.pt \
+            $lang_dir/L_disambig.fst
+    fi
+fi
+
+if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
+    log "Stage 6: Prepare BPE based lang"
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        lang_dir=data/lang_bpe_${vocab_size}
+        mkdir -p $lang_dir
+        # We reuse words.txt from phone based lexicon
+        # so that the two can share G.pt later.
+        cp data/lang_phone/words.txt $lang_dir
+
+        if [ ! -f $lang_dir/transcript_words.txt ]; then
+            log "Generate data for BPE training"
+
+            cat ./data/local/train/text | cut -d " " -f 2- >$lang_dir/transcript_words.txt
+        fi
+
+        if [ ! -f $lang_dir/bpe.model ]; then
+            ./local/train_bpe_model.py \
+                --lang-dir $lang_dir \
+                --vocab-size $vocab_size \
+                --transcript $lang_dir/transcript_words.txt
+        fi
+
+        if [ ! -f $lang_dir/L_disambig.pt ]; then
+            ./local/prepare_lang_bpe.py --lang-dir $lang_dir
+
+            log "Validating $lang_dir/lexicon.txt"
+            ./local/validate_bpe_lexicon.py \
+                --lexicon $lang_dir/lexicon.txt \
+                --bpe-model $lang_dir/bpe.model
+        fi
+
+        if [ ! -f $lang_dir/L.fst ]; then
+            log "Converting L.pt to L.fst"
+            ./shared/convert-k2-to-openfst.py \
+                --olabels aux_labels \
+                $lang_dir/L.pt \
+                $lang_dir/L.fst
+        fi
+
+        if [ ! -f $lang_dir/L_disambig.fst ]; then
+            log "Converting L_disambig.pt to L_disambig.fst"
+            ./shared/convert-k2-to-openfst.py \
+                --olabels aux_labels \
+                $lang_dir/L_disambig.pt \
+                $lang_dir/L_disambig.fst
+        fi
+    done
+fi
+
+if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
+    log "Stage 7: Prepare bigram token-level P for MMI training"
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        lang_dir=data/lang_bpe_${vocab_size}
+
+        if [ ! -f $lang_dir/transcript_tokens.txt ]; then
+            ./local/convert_transcript_words_to_tokens.py \
+                --lexicon $lang_dir/lexicon.txt \
+                --transcript $lang_dir/transcript_words.txt \
+                --oov "<unk>" \
+                >$lang_dir/transcript_tokens.txt
+        fi
+
+        if [ ! -f $lang_dir/P.arpa ]; then
+            ./shared/make_kn_lm.py \
+                -ngram-order 2 \
+                -text $lang_dir/transcript_tokens.txt \
+                -lm $lang_dir/P.arpa
+        fi
+
+        if [ ! -f $lang_dir/P.fst.txt ]; then
+            python3 -m kaldilm \
+                --read-symbol-table="$lang_dir/tokens.txt" \
+                --disambig-symbol='#0' \
+                --max-order=2 \
+                $lang_dir/P.arpa >$lang_dir/P.fst.txt
+        fi
+    done
+fi
+
+if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
+    log "Stage 8: Prepare G"
+    lang_dir=data/lang_phone
+    # We assume you have install kaldilm, if not, please install
+    # it using: pip install kaldilm
+
+    mkdir -p data/lm
+    if [ ! -f data/lm/G_3_gram.fst.txt ]; then
+        # It is used in building HLG
+        ./shared/make_kn_lm.py \
+            -ngram-order 3 \
+            -text ${lang_dir}/input.txt \
+            -lm data/lm/3-gram.arpa
+                    python3 -m kaldilm \
+                        --read-symbol-table="data/lang_phone/words.txt" \
+                        --disambig-symbol='#0' \
+                        --max-order=3 \
+                        data/lm/3-gram.arpa >data/lm/G_3_gram.fst.txt
+    fi
+
+    if [ ! -f data/lm/G_4_gram.fst.txt ]; then
+        # It is used for LM rescoring
+        ./shared/make_kn_lm.py \
+            -ngram-order 4 \
+            -text ${lang_dir}/input.txt \
+            -lm data/lm/4-gram.arpa
+                    python3 -m kaldilm \
+                        --read-symbol-table="data/lang_phone/words.txt" \
+                        --disambig-symbol='#0' \
+                        --max-order=4 \
+                        data/lm/4-gram.arpa >data/lm/G_4_gram.fst.txt
+    fi
+fi
+
+if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
+    log "Stage 9: Compile HLG"
+    ./local/compile_hlg.py --lang-dir data/lang_phone
+
+    # Note If ./local/compile_hlg.py throws OOM,
+    # please switch to the following command
+    #
+    # ./local/compile_hlg_using_openfst.py --lang-dir data/lang_phone
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        lang_dir=data/lang_bpe_${vocab_size}
+        ./local/compile_hlg.py --lang-dir $lang_dir
+
+        # Note If ./local/compile_hlg.py throws OOM,
+        # please switch to the following command
+        #
+        # ./local/compile_hlg_using_openfst.py --lang-dir $lang_dir
+    done
+fi
+
+# Compile LG for RNN-T fast_beam_search decoding
+if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
+    log "Stage 10: Compile LG"
+    ./local/compile_lg.py --lang-dir data/lang_phone
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        lang_dir=data/lang_bpe_${vocab_size}
+        ./local/compile_lg.py --lang-dir $lang_dir
+    done
+fi
+
+if [ $stage -le 11 ] && [ $stop_stage -ge 11 ]; then
+    log "Stage 11: Generate LM training data"
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        log "Processing vocab_size == ${vocab_size}"
+        lang_dir=data/lang_bpe_${vocab_size}
+        out_dir=data/lm_training_bpe_${vocab_size}
+        mkdir -p $out_dir
+
+        ./local/prepare_lm_training_data.py \
+            --bpe-model $lang_dir/bpe.model \
+            --lm-data data/lang_phone/input.txt \
+            --lm-archive $out_dir/lm_data.pt
+        done
+fi
+
+# if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
+#     log "Stage 12: Generate LM validation data"
+
+#     for vocab_size in ${vocab_sizes[@]}; do
+#         log "Processing vocab_size == ${vocab_size}"
+#         out_dir=data/lm_training_bpe_${vocab_size}
+#         mkdir -p $out_dir
+
+#         if [ ! -f $out_dir/valid.txt ]; then
+#             files=$(
+#                 find "$dl_dir/LibriSpeech/dev-clean" -name "*.trans.txt"
+#                 find "$dl_dir/LibriSpeech/dev-other" -name "*.trans.txt"
+#             )
+#             for f in ${files[@]}; do
+#                 cat $f | cut -d " " -f 2-
+#             done >$out_dir/valid.txt
+#         fi
+
+#         lang_dir=data/lang_bpe_${vocab_size}
+#         ./local/prepare_lm_training_data.py \
+#             --bpe-model $lang_dir/bpe.model \
+#             --lm-data $out_dir/valid.txt \
+#             --lm-archive $out_dir/lm_data-valid.pt
+#     done
+# fi
+
+if [ $stage -le 13 ] && [ $stop_stage -ge 13 ]; then
+    log "Stage 13: Generate LM test data"
+    testsets=(eval2000 rt03)
+
+    for testset in ${testsets[@]}; do
+        for vocab_size in ${vocab_sizes[@]}; do
+            log "Processing vocab_size == ${vocab_size}"
+            out_dir=data/lm_training_bpe_${vocab_size}
+            mkdir -p $out_dir
+
+            if [ ! -f $out_dir/test.txt ]; then
+                cat data/local/${testset}/text | cut -d " " -f 2- >$out_dir/${testset}.txt
+            fi
+
+            lang_dir=data/lang_bpe_${vocab_size}
+            ./local/prepare_lm_training_data.py \
+                --bpe-model $lang_dir/bpe.model \
+                --lm-data $out_dir/${testset}.txt \
+                --lm-archive $out_dir/lm_data-${testset}.pt
+            done
+        done
+fi
+
+if [ $stage -le 14 ] && [ $stop_stage -ge 14 ]; then
+    log "Stage 14: Sort LM training data"
+    testsets=(eval2000 rt03)
+    # Sort LM training data by sentence length in descending order
+    # for ease of training.
+    #
+    # Sentence length equals to the number of BPE tokens
+    # in a sentence.
+
+    for vocab_size in ${vocab_sizes[@]}; do
+        out_dir=data/lm_training_bpe_${vocab_size}
+        mkdir -p $out_dir
+        ./local/sort_lm_training_data.py \
+            --in-lm-data $out_dir/lm_data.pt \
+            --out-lm-data $out_dir/sorted_lm_data.pt \
+            --out-statistics $out_dir/statistics.txt
+                    for testset in ${testsets[@]}; do
+                        ./local/sort_lm_training_data.py \
+                            --in-lm-data $out_dir/lm_data-${testset}.pt \
+                            --out-lm-data $out_dir/sorted_lm_data-${testset}.pt \
+                            --out-statistics $out_dir/statistics-test-${testset}.txt
+                        done
+                    done
+fi

egs/swbd/ASR/shared

@@ -0,0 +1 @@
+../../../icefall/shared/

egs/swbd/ASR/utils/filter_scp.pl

@@ -0,0 +1,87 @@
+#!/usr/bin/env perl
+# Copyright 2010-2012 Microsoft Corporation
+#                     Johns Hopkins University (author: Daniel Povey)
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#  http://www.apache.org/licenses/LICENSE-2.0
+#
+# THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
+# WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
+# MERCHANTABLITY OR NON-INFRINGEMENT.
+# See the Apache 2 License for the specific language governing permissions and
+# limitations under the License.
+
+
+# This script takes a list of utterance-ids or any file whose first field
+# of each line is an utterance-id, and filters an scp
+# file (or any file whose "n-th" field is an utterance id), printing
+# out only those lines whose "n-th" field is in id_list. The index of
+# the "n-th" field is 1, by default, but can be changed by using
+# the -f <n> switch
+
+$exclude = 0;
+$field = 1;
+$shifted = 0;
+
+do {
+  $shifted=0;
+  if ($ARGV[0] eq "--exclude") {
+    $exclude = 1;
+    shift @ARGV;
+    $shifted=1;
+  }
+  if ($ARGV[0] eq "-f") {
+    $field = $ARGV[1];
+    shift @ARGV; shift @ARGV;
+    $shifted=1
+  }
+} while ($shifted);
+
+if(@ARGV < 1 || @ARGV > 2) {
+  die "Usage: filter_scp.pl [--exclude] [-f <field-to-filter-on>] id_list [in.scp] > out.scp \n" .
+      "Prints only the input lines whose f'th field (default: first) is in 'id_list'.\n" .
+      "Note: only the first field of each line in id_list matters.  With --exclude, prints\n" .
+      "only the lines that were *not* in id_list.\n" .
+      "Caution: previously, the -f option was interpreted as a zero-based field index.\n" .
+      "If your older scripts (written before Oct 2014) stopped working and you used the\n" .
+      "-f option, add 1 to the argument.\n" .
+      "See also: utils/filter_scp.pl .\n";
+}
+
+
+$idlist = shift @ARGV;
+open(F, "<$idlist") || die "Could not open id-list file $idlist";
+while(<F>) {
+  @A = split;
+  @A>=1 || die "Invalid id-list file line $_";
+  $seen{$A[0]} = 1;
+}
+
+if ($field == 1) { # Treat this as special case, since it is common.
+  while(<>) {
+    $_ =~ m/\s*(\S+)\s*/ || die "Bad line $_, could not get first field.";
+    # $1 is what we filter on.
+    if ((!$exclude && $seen{$1}) || ($exclude && !defined $seen{$1})) {
+      print $_;
+    }
+  }
+} else {
+  while(<>) {
+    @A = split;
+    @A > 0 || die "Invalid scp file line $_";
+    @A >= $field || die "Invalid scp file line $_";
+    if ((!$exclude && $seen{$A[$field-1]}) || ($exclude && !defined $seen{$A[$field-1]})) {
+      print $_;
+    }
+  }
+}
+
+# tests:
+# the following should print "foo 1"
+# ( echo foo 1; echo bar 2 ) | utils/filter_scp.pl <(echo foo)
+# the following should print "bar 2".
+# ( echo foo 1; echo bar 2 ) | utils/filter_scp.pl -f 2 <(echo 2)

egs/swbd/ASR/utils/fix_data_dir.sh

@@ -0,0 +1,197 @@
+#!/bin/bash
+
+# This script makes sure that only the segments present in
+# all of "feats.scp", "wav.scp" [if present], segments [if present]
+# text, and utt2spk are present in any of them.
+# It puts the original contents of data-dir into
+# data-dir/.backup
+
+cmd="$@"
+
+utt_extra_files=
+spk_extra_files=
+
+. utils/parse_options.sh
+
+if [ $# != 1 ]; then
+  echo "Usage: utils/data/fix_data_dir.sh <data-dir>"
+  echo "e.g.: utils/data/fix_data_dir.sh data/train"
+  echo "This script helps ensure that the various files in a data directory"
+  echo "are correctly sorted and filtered, for example removing utterances"
+  echo "that have no features (if feats.scp is present)"
+  exit 1
+fi
+
+data=$1
+
+if [ -f $data/images.scp ]; then
+  image/fix_data_dir.sh $cmd
+  exit $?
+fi
+
+mkdir -p $data/.backup
+
+[ ! -d $data ] && echo "$0: no such directory $data" && exit 1;
+
+[ ! -f $data/utt2spk ] && echo "$0: no such file $data/utt2spk" && exit 1;
+
+set -e -o pipefail -u
+
+tmpdir=$(mktemp -d /tmp/kaldi.XXXX);
+trap 'rm -rf "$tmpdir"' EXIT HUP INT PIPE TERM
+
+export LC_ALL=C
+
+function check_sorted {
+  file=$1
+  sort -k1,1 -u <$file >$file.tmp
+  if ! cmp -s $file $file.tmp; then
+    echo "$0: file $1 is not in sorted order or not unique, sorting it"
+    mv $file.tmp $file
+  else
+    rm $file.tmp
+  fi
+}
+
+for x in utt2spk spk2utt feats.scp text segments wav.scp cmvn.scp vad.scp \
+    reco2file_and_channel spk2gender utt2lang utt2uniq utt2dur reco2dur utt2num_frames; do
+  if [ -f $data/$x ]; then
+    cp $data/$x $data/.backup/$x
+    check_sorted $data/$x
+  fi
+done
+
+
+function filter_file {
+  filter=$1
+  file_to_filter=$2
+  cp $file_to_filter ${file_to_filter}.tmp
+  utils/filter_scp.pl $filter ${file_to_filter}.tmp > $file_to_filter
+  if ! cmp ${file_to_filter}.tmp  $file_to_filter >&/dev/null; then
+    length1=$(cat ${file_to_filter}.tmp | wc -l)
+    length2=$(cat ${file_to_filter} | wc -l)
+    if [ $length1 -ne $length2 ]; then
+      echo "$0: filtered $file_to_filter from $length1 to $length2 lines based on filter $filter."
+    fi
+  fi
+  rm $file_to_filter.tmp
+}
+
+function filter_recordings {
+  # We call this once before the stage when we filter on utterance-id, and once
+  # after.
+
+  if [ -f $data/segments ]; then
+  # We have a segments file -> we need to filter this and the file wav.scp, and
+  # reco2file_and_utt, if it exists, to make sure they have the same list of
+  # recording-ids.
+
+    if [ ! -f $data/wav.scp ]; then
+      echo "$0: $data/segments exists but not $data/wav.scp"
+      exit 1;
+    fi
+    awk '{print $2}' < $data/segments | sort | uniq > $tmpdir/recordings
+    n1=$(cat $tmpdir/recordings | wc -l)
+    [ ! -s $tmpdir/recordings ] && \
+      echo "Empty list of recordings (bad file $data/segments)?" && exit 1;
+    utils/filter_scp.pl $data/wav.scp $tmpdir/recordings > $tmpdir/recordings.tmp
+    mv $tmpdir/recordings.tmp $tmpdir/recordings
+
+
+    cp $data/segments{,.tmp}; awk '{print $2, $1, $3, $4}' <$data/segments.tmp >$data/segments
+    filter_file $tmpdir/recordings $data/segments
+    cp $data/segments{,.tmp}; awk '{print $2, $1, $3, $4}' <$data/segments.tmp >$data/segments
+    rm $data/segments.tmp
+
+    filter_file $tmpdir/recordings $data/wav.scp
+    [ -f $data/reco2file_and_channel ] && filter_file $tmpdir/recordings $data/reco2file_and_channel
+    [ -f $data/reco2dur ] && filter_file $tmpdir/recordings $data/reco2dur
+    true
+  fi
+}
+
+function filter_speakers {
+  # throughout this program, we regard utt2spk as primary and spk2utt as derived, so...
+  utils/utt2spk_to_spk2utt.pl $data/utt2spk > $data/spk2utt
+
+  cat $data/spk2utt | awk '{print $1}' > $tmpdir/speakers
+  for s in cmvn.scp spk2gender; do
+    f=$data/$s
+    if [ -f $f ]; then
+      filter_file $f $tmpdir/speakers
+    fi
+  done
+
+  filter_file $tmpdir/speakers $data/spk2utt
+  utils/spk2utt_to_utt2spk.pl $data/spk2utt > $data/utt2spk
+
+  for s in cmvn.scp spk2gender $spk_extra_files; do
+    f=$data/$s
+    if [ -f $f ]; then
+      filter_file $tmpdir/speakers $f
+    fi
+  done
+}
+
+function filter_utts {
+  cat $data/utt2spk | awk '{print $1}' > $tmpdir/utts
+
+  ! cat $data/utt2spk | sort | cmp - $data/utt2spk && \
+    echo "utt2spk is not in sorted order (fix this yourself)" && exit 1;
+
+  ! cat $data/utt2spk | sort -k2 | cmp - $data/utt2spk && \
+    echo "utt2spk is not in sorted order when sorted first on speaker-id " && \
+    echo "(fix this by making speaker-ids prefixes of utt-ids)" && exit 1;
+
+  ! cat $data/spk2utt | sort | cmp - $data/spk2utt && \
+    echo "spk2utt is not in sorted order (fix this yourself)" && exit 1;
+
+  if [ -f $data/utt2uniq ]; then
+    ! cat $data/utt2uniq | sort | cmp - $data/utt2uniq && \
+      echo "utt2uniq is not in sorted order (fix this yourself)" && exit 1;
+  fi
+
+  maybe_wav=
+  maybe_reco2dur=
+  [ ! -f $data/segments ] && maybe_wav=wav.scp # wav indexed by utts only if segments does not exist.
+  [ -s $data/reco2dur ] && [ ! -f $data/segments ] && maybe_reco2dur=reco2dur # reco2dur indexed by utts
+  for x in feats.scp text segments utt2lang $maybe_wav; do
+    if [ -f $data/$x ]; then
+      utils/filter_scp.pl $data/$x $tmpdir/utts > $tmpdir/utts.tmp
+      mv $tmpdir/utts.tmp $tmpdir/utts
+    fi
+  done
+  [ ! -s $tmpdir/utts ] && echo "fix_data_dir.sh: no utterances remained: not proceeding further." && \
+    rm $tmpdir/utts && exit 1;
+
+
+  if [ -f $data/utt2spk ]; then
+    new_nutts=$(cat $tmpdir/utts | wc -l)
+    old_nutts=$(cat $data/utt2spk | wc -l)
+    if [ $new_nutts -ne $old_nutts ]; then
+      echo "fix_data_dir.sh: kept $new_nutts utterances out of $old_nutts"
+    else
+      echo "fix_data_dir.sh: kept all $old_nutts utterances."
+    fi
+  fi
+
+  for x in utt2spk utt2uniq feats.scp vad.scp text segments utt2lang utt2dur utt2num_frames $maybe_wav $maybe_reco2dur $utt_extra_files; do
+    if [ -f $data/$x ]; then
+      cp $data/$x $data/.backup/$x
+      if ! cmp -s $data/$x <( utils/filter_scp.pl $tmpdir/utts $data/$x ) ; then
+        utils/filter_scp.pl $tmpdir/utts $data/.backup/$x > $data/$x
+      fi
+    fi
+  done
+
+}
+
+filter_recordings
+filter_speakers
+filter_utts
+filter_speakers
+filter_recordings
+
+utils/utt2spk_to_spk2utt.pl $data/utt2spk > $data/spk2utt
+
+echo "fix_data_dir.sh: old files are kept in $data/.backup"

egs/swbd/ASR/utils/parse_options.sh

@@ -0,0 +1,97 @@
+#!/bin/bash
+
+# Copyright 2012  Johns Hopkins University (Author: Daniel Povey);
+#                 Arnab Ghoshal, Karel Vesely
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#  http://www.apache.org/licenses/LICENSE-2.0
+#
+# THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
+# WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
+# MERCHANTABLITY OR NON-INFRINGEMENT.
+# See the Apache 2 License for the specific language governing permissions and
+# limitations under the License.
+
+
+# Parse command-line options.
+# To be sourced by another script (as in ". parse_options.sh").
+# Option format is: --option-name arg
+# and shell variable "option_name" gets set to value "arg."
+# The exception is --help, which takes no arguments, but prints the
+# $help_message variable (if defined).
+
+
+###
+### The --config file options have lower priority to command line
+### options, so we need to import them first...
+###
+
+# Now import all the configs specified by command-line, in left-to-right order
+for ((argpos=1; argpos<$#; argpos++)); do
+  if [ "${!argpos}" == "--config" ]; then
+    argpos_plus1=$((argpos+1))
+    config=${!argpos_plus1}
+    [ ! -r $config ] && echo "$0: missing config '$config'" && exit 1
+    . $config  # source the config file.
+  fi
+done
+
+
+###
+### No we process the command line options
+###
+while true; do
+  [ -z "${1:-}" ] && break;  # break if there are no arguments
+  case "$1" in
+    # If the enclosing script is called with --help option, print the help
+    # message and exit.  Scripts should put help messages in $help_message
+    --help|-h) if [ -z "$help_message" ]; then echo "No help found." 1>&2;
+      else printf "$help_message\n" 1>&2 ; fi;
+      exit 0 ;;
+    --*=*) echo "$0: options to scripts must be of the form --name value, got '$1'"
+      exit 1 ;;
+    # If the first command-line argument begins with "--" (e.g. --foo-bar),
+    # then work out the variable name as $name, which will equal "foo_bar".
+    --*) name=`echo "$1" | sed s/^--// | sed s/-/_/g`;
+      # Next we test whether the variable in question is undefned-- if so it's
+      # an invalid option and we die.  Note: $0 evaluates to the name of the
+      # enclosing script.
+      # The test [ -z ${foo_bar+xxx} ] will return true if the variable foo_bar
+      # is undefined.  We then have to wrap this test inside "eval" because
+      # foo_bar is itself inside a variable ($name).
+      eval '[ -z "${'$name'+xxx}" ]' && echo "$0: invalid option $1" 1>&2 && exit 1;
+
+      oldval="`eval echo \\$$name`";
+      # Work out whether we seem to be expecting a Boolean argument.
+      if [ "$oldval" == "true" ] || [ "$oldval" == "false" ]; then
+        was_bool=true;
+      else
+        was_bool=false;
+      fi
+
+      # Set the variable to the right value-- the escaped quotes make it work if
+      # the option had spaces, like --cmd "queue.pl -sync y"
+      eval $name=\"$2\";
+
+      # Check that Boolean-valued arguments are really Boolean.
+      if $was_bool && [[ "$2" != "true" && "$2" != "false" ]]; then
+        echo "$0: expected \"true\" or \"false\": $1 $2" 1>&2
+        exit 1;
+      fi
+      shift 2;
+      ;;
+  *) break;
+  esac
+done
+
+
+# Check for an empty argument to the --cmd option, which can easily occur as a
+# result of scripting errors.
+[ ! -z "${cmd+xxx}" ] && [ -z "$cmd" ] && echo "$0: empty argument to --cmd option" 1>&2 && exit 1;
+
+
+true; # so this script returns exit code 0.

egs/swbd/ASR/utils/spk2utt_to_utt2spk.pl

@@ -0,0 +1,27 @@
+#!/usr/bin/env perl
+# Copyright 2010-2011 Microsoft Corporation
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#  http://www.apache.org/licenses/LICENSE-2.0
+#
+# THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
+# WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
+# MERCHANTABLITY OR NON-INFRINGEMENT.
+# See the Apache 2 License for the specific language governing permissions and
+# limitations under the License.
+
+
+while(<>){ 
+    @A = split(" ", $_);
+    @A > 1 || die "Invalid line in spk2utt file: $_";
+    $s = shift @A;
+    foreach $u ( @A ) {
+        print "$u $s\n";
+    }
+}
+
+

egs/swbd/ASR/utils/utt2spk_to_spk2utt.pl

@@ -0,0 +1,38 @@
+#!/usr/bin/env perl
+# Copyright 2010-2011 Microsoft Corporation
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#  http://www.apache.org/licenses/LICENSE-2.0
+#
+# THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
+# WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
+# MERCHANTABLITY OR NON-INFRINGEMENT.
+# See the Apache 2 License for the specific language governing permissions and
+# limitations under the License.
+
+# converts an utt2spk file to a spk2utt file.
+# Takes input from the stdin or from a file argument;
+# output goes to the standard out.
+
+if ( @ARGV > 1 ) {
+    die "Usage: utt2spk_to_spk2utt.pl [ utt2spk ] > spk2utt";
+}
+
+while(<>){ 
+    @A = split(" ", $_);
+    @A == 2 || die "Invalid line in utt2spk file: $_";
+    ($u,$s) = @A;
+    if(!$seen_spk{$s}) {
+        $seen_spk{$s} = 1;
+        push @spklist, $s;
+    }
+    push (@{$spk_hash{$s}}, "$u");
+}
+foreach $s (@spklist) {
+    $l = join(' ',@{$spk_hash{$s}});
+    print "$s $l\n";
+}