Add Streaming Zipformer-Transducer recipe for KsponSpeech (#1651)

egs/ksponspeech/ASR/README.md

@@ -0,0 +1,32 @@
+# Introduction
+KsponSpeech is a large-scale spontaneous speech corpus of Korean.
+This corpus contains 969 hours of open-domain dialog utterances,
+spoken by about 2,000 native Korean speakers in a clean environment.
+
+All data were constructed by recording the dialogue of two people
+freely conversing on a variety of topics and manually transcribing the utterances.
+
+The transcription provides a dual transcription consisting of orthography and pronunciation,
+and disfluency tags for spontaneity of speech, such as filler words, repeated words, and word fragments.
+
+The original audio data has a pcm extension.
+During preprocessing, it is converted into a file in the flac extension and saved anew.
+
+KsponSpeech is publicly available on an open data hub site of the Korea government.
+The dataset must be downloaded manually.
+
+For more details, please visit:
+
+ - Dataset: https://aihub.or.kr/aihubdata/data/view.do?currMenu=115&topMenu=100&aihubDataSe=realm&dataSetSn=123
+ - Paper: https://www.mdpi.com/2076-3417/10/19/6936
+
+[./RESULTS.md](./RESULTS.md) contains the latest results.
+
+# Transducers
+There are various folders containing the name `transducer` in this folder. The following table lists the differences among them.
+
+|                                          | Encoder              | Decoder            | Comment                                           |
+| ---------------------------------------- | -------------------- | ------------------ | ------------------------------------------------- |
+| `pruned_transducer_stateless7_streaming`                              | Streaming Zipformer   | Embedding + Conv1d | streaming version of pruned_transducer_stateless7                                 |
+
+The decoder in `transducer_stateless` is modified from the paper [Rnn-Transducer with Stateless Prediction Network](https://ieeexplore.ieee.org/document/9054419/). We place an additional Conv1d layer right after the input embedding layer.

egs/ksponspeech/ASR/RESULTS.md

@@ -0,0 +1,70 @@
+## Results
+
+### Streaming Zipformer-Transducer (Pruned Stateless Transducer + Streaming Zipformer)
+
+#### [pruned_transducer_stateless7_streaming](./pruned_transducer_stateless7_streaming)
+
+Number of model parameters: 79,022,891, i.e., 79.02 M
+
+##### Training on KsponSpeech (with MUSAN)
+
+Model: [johnBamma/icefall-asr-ksponspeech-pruned-transducer-stateless7-streaming-2024-06-12](https://huggingface.co/johnBamma/icefall-asr-ksponspeech-pruned-transducer-stateless7-streaming-2024-06-12)
+
+The CERs are:
+
+| decoding method      | chunk size | eval_clean | eval_other | comment             | decoding mode        |
+|----------------------|------------|------------|------------|---------------------|----------------------|
+| greedy search        | 320ms      | 10.21      | 11.07      | --epoch 30 --avg 9  | simulated streaming  |
+| greedy search        | 320ms      | 10.22      | 11.07      | --epoch 30 --avg 9  | chunk-wise           |
+| fast beam search     | 320ms      | 10.21      | 11.04      | --epoch 30 --avg 9  | simulated streaming  |
+| fast beam search     | 320ms      | 10.25      | 11.08      | --epoch 30 --avg 9  | chunk-wise           |
+| modified beam search | 320ms      | 10.13      | 10.88      | --epoch 30 --avg 9  | simulated streaming  |
+| modified beam search | 320ms      | 10.1       | 10.93      | --epoch 30 --avg 9  | chunk-size           |
+| greedy search        | 640ms      | 9.94       | 10.82      | --epoch 30 --avg 9  | simulated streaming  |
+| greedy search        | 640ms      | 10.04      | 10.85      | --epoch 30 --avg 9  | chunk-wise           |
+| fast beam search     | 640ms      | 10.01      | 10.81      | --epoch 30 --avg 9  | simulated streaming  |
+| fast beam search     | 640ms      | 10.04      | 10.7       | --epoch 30 --avg 9  | chunk-wise           |
+| modified beam search | 640ms      | 9.91       | 10.72      | --epoch 30 --avg 9  | simulated streaming  |
+| modified beam search | 640ms      | 9.92       | 10.72      | --epoch 30 --avg 9  | chunk-size           |
+
+Note: `simulated streaming` indicates feeding full utterance during decoding using `decode.py`,
+while `chunk-size` indicates feeding certain number of frames at each time using `streaming_decode.py`.
+
+The training command is:
+
+```bash
+./pruned_transducer_stateless7_streaming/train.py \
+    --world-size 4 \
+    --num-epochs 30 \
+    --start-epoch 1 \
+    --use-fp16 1 \
+    --exp-dir pruned_transducer_stateless7_streaming/exp \
+    --max-duration 750 \
+    --enable-musan True
+```
+
+The simulated streaming decoding command (e.g., chunk-size=320ms) is:
+```bash
+for m in greedy_search fast_beam_search modified_beam_search; do
+  ./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 30 \
+    --avg 9 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method $m
+done
+```
+
+The streaming chunk-size decoding command (e.g., chunk-size=320ms) is:
+```bash
+for m in greedy_search modified_beam_search fast_beam_search; do
+  ./pruned_transducer_stateless7_streaming/streaming_decode.py \
+    --epoch 30 \
+    --avg 9 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --decoding-method $m \
+    --decode-chunk-len 32 \
+    --num-decode-streams 2000
+done
+```

egs/ksponspeech/ASR/local/compute_fbank_ksponspeech.py

@@ -0,0 +1,185 @@
+#!/usr/bin/env python3
+# Copyright    2024   (Author: SeungHyun Lee, Contacts: whsqkaak@naver.com)
+#
+# 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 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.""",
+    )
+    parser.add_argument(
+        "--data-dir",
+        type=str,
+        default="data",
+        help="""Path of data directory""",
+    )
+
+    return parser.parse_args()
+
+
+def compute_fbank_speechtools(
+    bpe_model: Optional[str] = None,
+    dataset: Optional[str] = None,
+    perturb_speed: Optional[bool] = False,
+    data_dir: Optional[str] = "data",
+):
+    src_dir = Path(data_dir) / "manifests"
+    output_dir = Path(data_dir) / "fbank"
+    num_jobs = min(4, 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 = (
+            "train",
+            "dev",
+            "eval_clean",
+            "eval_other",
+        )
+    else:
+        dataset_parts = dataset.split(" ", -1)
+
+    prefix = "ksponspeech"
+    suffix = "jsonl.gz"
+    logging.info(f"Read manifests...")
+    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,
+    )
+
+    if torch.cuda.is_available():
+        # Use cuda for fbank compute
+        device = "cuda"
+    else:
+        device = "cpu"
+    logging.info(f"Device: {device}")
+
+    extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins, device=device))
+
+    with get_executor() as ex:  # Initialize the executor only once.
+        logging.info(f"Executor: {ex}")
+        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"],
+                supervisions=m["supervisions"],
+            )
+
+            # Filter duration
+            cut_set = cut_set.filter(
+                lambda x: x.duration > 1 and x.sampling_rate == 16000
+            )
+
+            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)
+                    )
+            logging.info(f"Compute & Store features...")
+            if device == "cuda":
+                cut_set = cut_set.compute_and_store_features_batch(
+                    extractor=extractor,
+                    storage_path=f"{output_dir}/{prefix}_feats_{partition}",
+                    num_workers=4,
+                    storage_type=LilcomChunkyWriter,
+                )
+            else:
+                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_speechtools(
+        bpe_model=args.bpe_model,
+        dataset=args.dataset,
+        perturb_speed=args.perturb_speed,
+        data_dir=args.data_dir,
+    )

egs/ksponspeech/ASR/local/compute_fbank_musan.py

@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+# 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 musan dataset.
+It looks for manifests in the directory `src_dir` (default is data/manifests).
+
+The generated fbank features are saved in data/fbank.
+"""
+import argparse
+import logging
+import os
+from pathlib import Path
+
+import torch
+from lhotse import (
+    CutSet,
+    Fbank,
+    FbankConfig,
+    LilcomChunkyWriter,
+    MonoCut,
+    WhisperFbank,
+    WhisperFbankConfig,
+    combine,
+)
+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 is_cut_long(c: MonoCut) -> bool:
+    return c.duration > 5
+
+
+def compute_fbank_musan(
+    src_dir: str = "data/manifests",
+    num_mel_bins: int = 80,
+    whisper_fbank: bool = False,
+    output_dir: str = "data/fbank",
+):
+    src_dir = Path(src_dir)
+    output_dir = Path(output_dir)
+    num_jobs = min(15, os.cpu_count())
+
+    dataset_parts = (
+        "music",
+        "speech",
+        "noise",
+    )
+    prefix = "musan"
+    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,
+    )
+
+    musan_cuts_path = output_dir / "musan_cuts.jsonl.gz"
+
+    if musan_cuts_path.is_file():
+        logging.info(f"{musan_cuts_path} already exists - skipping")
+        return
+
+    logging.info("Extracting features for Musan")
+
+    if whisper_fbank:
+        extractor = WhisperFbank(
+            WhisperFbankConfig(num_filters=num_mel_bins, device="cuda")
+        )
+    else:
+        extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
+
+    with get_executor() as ex:  # Initialize the executor only once.
+        # create chunks of Musan with duration 5 - 10 seconds
+        musan_cuts = (
+            CutSet.from_manifests(
+                recordings=combine(part["recordings"] for part in manifests.values())
+            )
+            .cut_into_windows(10.0)
+            .filter(is_cut_long)
+            .compute_and_store_features(
+                extractor=extractor,
+                storage_path=f"{output_dir}/musan_feats",
+                num_jobs=num_jobs if ex is None else 80,
+                executor=ex,
+                storage_type=LilcomChunkyWriter,
+            )
+        )
+        musan_cuts.to_file(musan_cuts_path)
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--src-dir",
+        type=str,
+        default="data/manifests",
+        help="Source manifests directory.",
+    )
+    parser.add_argument(
+        "--num-mel-bins",
+        type=int,
+        default=80,
+        help="""The number of mel bins for Fbank""",
+    )
+    parser.add_argument(
+        "--whisper-fbank",
+        type=str2bool,
+        default=False,
+        help="Use WhisperFbank instead of Fbank. Default: False.",
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        default="data/fbank",
+        help="Output directory. Default: data/fbank.",
+    )
+    return parser.parse_args()
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    args = get_args()
+    compute_fbank_musan(
+        src_dir=args.src_dir,
+        num_mel_bins=args.num_mel_bins,
+        whisper_fbank=args.whisper_fbank,
+        output_dir=args.output_dir,
+    )

egs/ksponspeech/ASR/local/filter_cuts.py

@@ -0,0 +1,157 @@
+#!/usr/bin/env python3
+
+# 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_5000/bpe.model \
+    --in-cuts data/fbank/speechtools_cuts_test.jsonl.gz \
+    --out-cuts data/fbank-filtered/speechtools_cuts_test.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/ksponspeech/ASR/local/train_bpe_model.py

@@ -0,0 +1,115 @@
+#!/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
+from typing import Dict
+
+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 generate_tokens(lang_dir: Path):
+    """
+    Generate the tokens.txt from a bpe model.
+    """
+    sp = spm.SentencePieceProcessor()
+    sp.load(str(lang_dir / "bpe.model"))
+    token2id: Dict[str, int] = {sp.id_to_piece(i): i for i in range(sp.vocab_size())}
+    with open(lang_dir / "tokens.txt", "w", encoding="utf-8") as f:
+        for sym, i in token2id.items():
+            f.write(f"{sym} {i}\n")
+
+
+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")
+
+    generate_tokens(lang_dir)
+
+
+if __name__ == "__main__":
+    main()

egs/ksponspeech/ASR/local/validate_manifest.py

@@ -0,0 +1,101 @@
+#!/usr/bin/env python3
+# 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 the following assumptions of the generated manifest:
+
+- Single supervision per cut
+- Supervision time bounds are within cut time bounds
+
+We will add more checks later if needed.
+
+Usage example:
+
+    python3 ./local/validate_manifest.py \
+            ./data/fbank/speechtools_cuts_train.jsonl.gz
+
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+from lhotse import CutSet, load_manifest_lazy
+from lhotse.cut import Cut
+from lhotse.dataset.speech_recognition import validate_for_asr
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "manifest",
+        type=Path,
+        help="Path to the manifest file",
+    )
+
+    return parser.parse_args()
+
+
+def validate_one_supervision_per_cut(c: Cut):
+    if len(c.supervisions) != 1:
+        raise ValueError(f"{c.id} has {len(c.supervisions)} supervisions")
+
+
+def validate_supervision_and_cut_time_bounds(c: Cut):
+    tol = 2e-3  # same tolerance as in 'validate_for_asr()'
+    s = c.supervisions[0]
+
+    # Supervision start time is relative to Cut ...
+    # https://lhotse.readthedocs.io/en/v0.10_e/cuts.html
+    if s.start < -tol:
+        raise ValueError(
+            f"{c.id}: Supervision start time {s.start} must not be negative."
+        )
+    if s.start > tol:
+        raise ValueError(
+            f"{c.id}: Supervision start time {s.start} is not at the beginning of the Cut. Please apply `lhotse cut trim-to-supervisions`."
+        )
+    if c.start + s.end > c.end + tol:
+        raise ValueError(
+            f"{c.id}: Supervision end time {c.start+s.end} is larger "
+            f"than cut end time {c.end}"
+        )
+
+
+def main():
+    args = get_args()
+
+    manifest = args.manifest
+    logging.info(f"Validating {manifest}")
+
+    assert manifest.is_file(), f"{manifest} does not exist"
+    cut_set = load_manifest_lazy(manifest)
+    assert isinstance(cut_set, CutSet)
+
+    for c in cut_set:
+        validate_one_supervision_per_cut(c)
+        validate_supervision_and_cut_time_bounds(c)
+
+    # Validation from K2 training
+    # - checks supervision start is 0
+    # - checks supervision.duration is not longer than cut.duration
+    # - there is tolerance 2ms
+    validate_for_asr(cut_set)
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    main()

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/README.md

@@ -0,0 +1 @@
+This recipe implements Streaming Zipformer-Transducer model.

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/asr_datamodule.py

@@ -0,0 +1,415 @@
+# Copyright    2024   (Author: SeungHyun Lee, Contacts: whsqkaak@naver.com)
+#
+# 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,
+    SimpleCutSampler,
+    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 KsponSpeechAsrDataModule:
+    """
+    DataModule for k2 ASR experiments.
+    It assumes there is always one train and valid dataloader.
+
+    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, p=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,
+                buffer_size=self.args.num_buckets * 2000,
+                shuffle_buffer_size=self.args.num_buckets * 5000,
+                drop_last=self.args.drop_last,
+            )
+        else:
+            logging.info("Using SimpleCutSampler.")
+            train_sampler = SimpleCutSampler(
+                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_cuts(self) -> CutSet:
+        logging.info("About to get train cuts.")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "ksponspeech_cuts_train.jsonl.gz"
+        )
+
+    @lru_cache()
+    def dev_cuts(self) -> CutSet:
+        logging.info("About to get dev cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "ksponspeech_cuts_dev.jsonl.gz"
+        )
+
+    @lru_cache()
+    def eval_clean_cuts(self) -> CutSet:
+        logging.info("About to get eval_clean cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "ksponspeech_cuts_eval_clean.jsonl.gz"
+        )
+
+    @lru_cache()
+    def eval_other_cuts(self) -> CutSet:
+        logging.info("About to get eval_other cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "ksponspeech_cuts_eval_other.jsonl.gz"
+        )

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/beam_search.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/beam_search.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/decode.py

@@ -0,0 +1,993 @@
+#!/usr/bin/env python3
+# Copyright    2024   (Author: SeungHyun Lee, Contacts: whsqkaak@naver.com)
+#
+# 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:
+(1) greedy search
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method greedy_search
+
+(2) beam search (not recommended)
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method beam_search \
+    --beam-size 4
+
+(3) modified beam search
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method modified_beam_search \
+    --beam-size 4
+
+(4) fast beam search (one best)
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method fast_beam_search \
+    --beam 20.0 \
+    --max-contexts 8 \
+    --max-states 64
+
+(5) fast beam search (nbest)
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method fast_beam_search_nbest \
+    --beam 20.0 \
+    --max-contexts 8 \
+    --max-states 64 \
+    --num-paths 200 \
+    --nbest-scale 0.5
+
+(6) fast beam search (nbest oracle WER)
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method fast_beam_search_nbest_oracle \
+    --beam 20.0 \
+    --max-contexts 8 \
+    --max-states 64 \
+    --num-paths 200 \
+    --nbest-scale 0.5
+
+(7) fast beam search (with LG)
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method fast_beam_search_nbest_LG \
+    --beam 20.0 \
+    --max-contexts 8 \
+    --max-states 64
+"""
+
+
+import argparse
+import logging
+import math
+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 KsponSpeechAsrDataModule
+from beam_search import (
+    beam_search,
+    fast_beam_search_nbest,
+    fast_beam_search_nbest_LG,
+    fast_beam_search_nbest_oracle,
+    fast_beam_search_one_best,
+    greedy_search,
+    greedy_search_batch,
+    modified_beam_search,
+    modified_beam_search_lm_rescore,
+    modified_beam_search_lm_rescore_LODR,
+    modified_beam_search_lm_shallow_fusion,
+    modified_beam_search_LODR,
+)
+from train import add_model_arguments, get_params, get_transducer_model
+
+from icefall import LmScorer, NgramLm
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.lexicon import Lexicon
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+LOG_EPS = math.log(1e-10)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=30,
+        help="""It specifies the checkpoint to use for decoding.
+        Note: Epoch counts from 1.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=9,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="pruned_transducer_stateless7_streaming/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=Path,
+        default="data/lang_bpe_500",
+        help="The lang dir containing word table and LG graph",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+          - modified_beam_search
+          - fast_beam_search
+          - fast_beam_search_nbest
+          - fast_beam_search_nbest_oracle
+          - fast_beam_search_nbest_LG
+        If you use fast_beam_search_nbest_LG, you have to specify
+        `--lang-dir`, which should contain `LG.pt`.
+        """,
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=4,
+        help="""An integer indicating how many candidates we will keep for each
+        frame. Used only when --decoding-method is beam_search or
+        modified_beam_search.""",
+    )
+
+    parser.add_argument(
+        "--beam",
+        type=float,
+        default=20.0,
+        help="""A floating point value to calculate the cutoff score during beam
+        search (i.e., `cutoff = max-score - beam`), which is the same as the
+        `beam` in Kaldi.
+        Used only when --decoding-method is fast_beam_search,
+        fast_beam_search_nbest, fast_beam_search_nbest_LG,
+        and fast_beam_search_nbest_oracle
+        """,
+    )
+
+    parser.add_argument(
+        "--ngram-lm-scale",
+        type=float,
+        default=0.01,
+        help="""
+        Used only when --decoding_method is fast_beam_search_nbest_LG.
+        It specifies the scale for n-gram LM scores.
+        """,
+    )
+
+    parser.add_argument(
+        "--max-contexts",
+        type=int,
+        default=8,
+        help="""Used only when --decoding-method is
+        fast_beam_search, fast_beam_search_nbest, fast_beam_search_nbest_LG,
+        and fast_beam_search_nbest_oracle""",
+    )
+
+    parser.add_argument(
+        "--max-states",
+        type=int,
+        default=64,
+        help="""Used only when --decoding-method is
+        fast_beam_search, fast_beam_search_nbest, fast_beam_search_nbest_LG,
+        and fast_beam_search_nbest_oracle""",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+    parser.add_argument(
+        "--max-sym-per-frame",
+        type=int,
+        default=1,
+        help="""Maximum number of symbols per frame.
+        Used only when --decoding_method is greedy_search""",
+    )
+
+    parser.add_argument(
+        "--num-paths",
+        type=int,
+        default=200,
+        help="""Number of paths for nbest decoding.
+        Used only when the decoding method is fast_beam_search_nbest,
+        fast_beam_search_nbest_LG, and fast_beam_search_nbest_oracle""",
+    )
+
+    parser.add_argument(
+        "--nbest-scale",
+        type=float,
+        default=0.5,
+        help="""Scale applied to lattice scores when computing nbest paths.
+        Used only when the decoding method is fast_beam_search_nbest,
+        fast_beam_search_nbest_LG, and fast_beam_search_nbest_oracle""",
+    )
+
+    parser.add_argument(
+        "--use-shallow-fusion",
+        type=str2bool,
+        default=False,
+        help="""Use neural network LM for shallow fusion.
+        If you want to use LODR, you will also need to set this to true
+        """,
+    )
+
+    parser.add_argument(
+        "--lm-type",
+        type=str,
+        default="rnn",
+        help="Type of NN lm",
+        choices=["rnn", "transformer"],
+    )
+
+    parser.add_argument(
+        "--lm-scale",
+        type=float,
+        default=0.3,
+        help="""The scale of the neural network LM
+        Used only when `--use-shallow-fusion` is set to True.
+        """,
+    )
+
+    parser.add_argument(
+        "--tokens-ngram",
+        type=int,
+        default=2,
+        help="""The order of the ngram lm.
+        """,
+    )
+
+    parser.add_argument(
+        "--backoff-id",
+        type=int,
+        default=500,
+        help="ID of the backoff symbol in the ngram LM",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+    word_table: Optional[k2.SymbolTable] = None,
+    decoding_graph: Optional[k2.Fsa] = None,
+    LM: Optional[LmScorer] = None,
+    ngram_lm=None,
+    ngram_lm_scale: float = 0.0,
+) -> 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 greedy_search is used, it would be "greedy_search"
+               If beam search with a beam size of 7 is used, it would be
+               "beam_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`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      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.
+      decoding_graph:
+        The decoding graph. Can be either a `k2.trivial_graph` or HLG, Used
+        only when --decoding_method is fast_beam_search, fast_beam_search_nbest,
+        fast_beam_search_nbest_oracle, and fast_beam_search_nbest_LG.
+      LM:
+        A neural network language model.
+      ngram_lm:
+        A ngram language model
+      ngram_lm_scale:
+        The scale for the ngram language model.
+    Returns:
+      Return the decoding result. See above description for the format of
+      the returned dict.
+    """
+    device = next(model.parameters()).device
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+
+    feature = feature.to(device)
+    # at entry, feature is (N, T, C)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    feature_lens += 30
+    feature = torch.nn.functional.pad(
+        feature,
+        pad=(0, 0, 0, 30),
+        value=LOG_EPS,
+    )
+    encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
+
+    hyps = []
+
+    if params.decoding_method == "fast_beam_search":
+        hyp_tokens = fast_beam_search_one_best(
+            model=model,
+            decoding_graph=decoding_graph,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam,
+            max_contexts=params.max_contexts,
+            max_states=params.max_states,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "fast_beam_search_nbest_LG":
+        hyp_tokens = fast_beam_search_nbest_LG(
+            model=model,
+            decoding_graph=decoding_graph,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam,
+            max_contexts=params.max_contexts,
+            max_states=params.max_states,
+            num_paths=params.num_paths,
+            nbest_scale=params.nbest_scale,
+        )
+        for hyp in hyp_tokens:
+            hyps.append([word_table[i] for i in hyp])
+    elif params.decoding_method == "fast_beam_search_nbest":
+        hyp_tokens = fast_beam_search_nbest(
+            model=model,
+            decoding_graph=decoding_graph,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam,
+            max_contexts=params.max_contexts,
+            max_states=params.max_states,
+            num_paths=params.num_paths,
+            nbest_scale=params.nbest_scale,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "fast_beam_search_nbest_oracle":
+        hyp_tokens = fast_beam_search_nbest_oracle(
+            model=model,
+            decoding_graph=decoding_graph,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam,
+            max_contexts=params.max_contexts,
+            max_states=params.max_states,
+            num_paths=params.num_paths,
+            ref_texts=sp.encode(supervisions["text"]),
+            nbest_scale=params.nbest_scale,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "greedy_search" and params.max_sym_per_frame == 1:
+        hyp_tokens = greedy_search_batch(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search":
+        hyp_tokens = modified_beam_search(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search_lm_shallow_fusion":
+        hyp_tokens = modified_beam_search_lm_shallow_fusion(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LM=LM,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search_LODR":
+        hyp_tokens = modified_beam_search_LODR(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LODR_lm=ngram_lm,
+            LODR_lm_scale=ngram_lm_scale,
+            LM=LM,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search_lm_rescore":
+        lm_scale_list = [0.01 * i for i in range(10, 50)]
+        ans_dict = modified_beam_search_lm_rescore(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LM=LM,
+            lm_scale_list=lm_scale_list,
+        )
+    elif params.decoding_method == "modified_beam_search_lm_rescore_LODR":
+        lm_scale_list = [0.02 * i for i in range(2, 30)]
+        ans_dict = modified_beam_search_lm_rescore_LODR(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LM=LM,
+            LODR_lm=ngram_lm,
+            sp=sp,
+            lm_scale_list=lm_scale_list,
+        )
+    else:
+        batch_size = encoder_out.size(0)
+
+        for i in range(batch_size):
+            # fmt: off
+            encoder_out_i = encoder_out[i:i+1, :encoder_out_lens[i]]
+            # fmt: on
+            if params.decoding_method == "greedy_search":
+                hyp = greedy_search(
+                    model=model,
+                    encoder_out=encoder_out_i,
+                    max_sym_per_frame=params.max_sym_per_frame,
+                )
+            elif params.decoding_method == "beam_search":
+                hyp = beam_search(
+                    model=model,
+                    encoder_out=encoder_out_i,
+                    beam=params.beam_size,
+                )
+            else:
+                raise ValueError(
+                    f"Unsupported decoding method: {params.decoding_method}"
+                )
+            hyps.append(sp.decode(hyp).split())
+
+    if params.decoding_method == "greedy_search":
+        return {"greedy_search": hyps}
+    elif "fast_beam_search" in params.decoding_method:
+        key = f"beam_{params.beam}_"
+        key += f"max_contexts_{params.max_contexts}_"
+        key += f"max_states_{params.max_states}"
+        if "nbest" in params.decoding_method:
+            key += f"_num_paths_{params.num_paths}_"
+            key += f"nbest_scale_{params.nbest_scale}"
+            if "LG" in params.decoding_method:
+                key += f"_ngram_lm_scale_{params.ngram_lm_scale}"
+
+        return {key: hyps}
+    elif params.decoding_method in (
+        "modified_beam_search_lm_rescore",
+        "modified_beam_search_lm_rescore_LODR",
+    ):
+        ans = dict()
+        assert ans_dict is not None
+        for key, hyps in ans_dict.items():
+            hyps = [sp.decode(hyp).split() for hyp in hyps]
+            ans[f"beam_size_{params.beam_size}_{key}"] = hyps
+        return ans
+    else:
+        return {f"beam_size_{params.beam_size}": hyps}
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    word_table: Optional[k2.SymbolTable] = None,
+    decoding_graph: Optional[k2.Fsa] = None,
+    LM: Optional[LmScorer] = None,
+    ngram_lm=None,
+    ngram_lm_scale: float = 0.0,
+) -> 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.
+      sp:
+        The BPE model.
+      word_table:
+        The word symbol table.
+      decoding_graph:
+        The decoding graph. Can be either a `k2.trivial_graph` or HLG, Used
+        only when --decoding_method is fast_beam_search, fast_beam_search_nbest,
+        fast_beam_search_nbest_oracle, and fast_beam_search_nbest_LG.
+      ngram_lm:
+        A n-gram LM to be used for LODR.
+    Returns:
+      Return a dict, whose key may be "greedy_search" if greedy search
+      is used, or it may be "beam_7" if beam size of 7 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 = "?"
+
+    if params.decoding_method == "greedy_search":
+        log_interval = 50
+    else:
+        log_interval = 20
+
+    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,
+            sp=sp,
+            decoding_graph=decoding_graph,
+            word_table=word_table,
+            batch=batch,
+            LM=LM,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+        )
+
+        for name, 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[name].extend(this_batch)
+
+        num_cuts += len(texts)
+
+        if batch_idx % log_interval == 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[str], List[str]]]],
+):
+    test_set_cers = dict()
+    for key, results in results_dict.items():
+        recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out CERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
+        with open(errs_filename, "w") as f:
+            cer = write_error_stats(
+                f,
+                f"{test_set_name}-{key}",
+                results,
+                enable_log=True,
+                compute_CER=True,
+            )
+            test_set_cers[key] = cer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_cers = sorted(test_set_cers.items(), key=lambda x: x[1])
+    errs_info = params.res_dir / f"cer-summary-{test_set_name}-{params.suffix}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tCER", file=f)
+        for key, val in test_set_cers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, CER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_cers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    KsponSpeechAsrDataModule.add_arguments(parser)
+    LmScorer.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    assert params.decoding_method in (
+        "greedy_search",
+        "beam_search",
+        "fast_beam_search",
+        "fast_beam_search_nbest",
+        "fast_beam_search_nbest_LG",
+        "fast_beam_search_nbest_oracle",
+        "modified_beam_search",
+        "modified_beam_search_LODR",
+        "modified_beam_search_lm_shallow_fusion",
+        "modified_beam_search_lm_rescore",
+        "modified_beam_search_lm_rescore_LODR",
+    )
+    params.res_dir = params.exp_dir / params.decoding_method
+
+    if params.iter > 0:
+        params.suffix = f"iter-{params.iter}-avg-{params.avg}"
+    else:
+        params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+
+    params.suffix += f"-streaming-chunk-size-{params.decode_chunk_len}"
+
+    if "fast_beam_search" in params.decoding_method:
+        params.suffix += f"-beam-{params.beam}"
+        params.suffix += f"-max-contexts-{params.max_contexts}"
+        params.suffix += f"-max-states-{params.max_states}"
+        if "nbest" in params.decoding_method:
+            params.suffix += f"-nbest-scale-{params.nbest_scale}"
+            params.suffix += f"-num-paths-{params.num_paths}"
+            if "LG" in params.decoding_method:
+                params.suffix += f"-ngram-lm-scale-{params.ngram_lm_scale}"
+    elif "beam_search" in params.decoding_method:
+        params.suffix += f"-{params.decoding_method}-beam-size-{params.beam_size}"
+    else:
+        params.suffix += f"-context-{params.context_size}"
+        params.suffix += f"-max-sym-per-frame-{params.max_sym_per_frame}"
+
+    if params.use_shallow_fusion:
+        params.suffix += f"-{params.lm_type}-lm-scale-{params.lm_scale}"
+
+        if "LODR" in params.decoding_method:
+            params.suffix += (
+                f"-LODR-{params.tokens_ngram}gram-scale-{params.ngram_lm_scale}"
+            )
+
+    if params.use_averaged_model:
+        params.suffix += "-use-averaged-model"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <unk> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.unk_id = sp.piece_to_id("<unk>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+    assert model.encoder.decode_chunk_size == params.decode_chunk_len // 2, (
+        model.encoder.decode_chunk_size,
+        params.decode_chunk_len,
+    )
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif 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 i >= 1:
+                    filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.to(device)
+    model.eval()
+
+    # only load the neural network LM if required
+    if params.use_shallow_fusion or params.decoding_method in (
+        "modified_beam_search_lm_rescore",
+        "modified_beam_search_lm_rescore_LODR",
+        "modified_beam_search_lm_shallow_fusion",
+        "modified_beam_search_LODR",
+    ):
+        LM = LmScorer(
+            lm_type=params.lm_type,
+            params=params,
+            device=device,
+            lm_scale=params.lm_scale,
+        )
+        LM.to(device)
+        LM.eval()
+    else:
+        LM = None
+
+    # only load N-gram LM when needed
+    if params.decoding_method == "modified_beam_search_lm_rescore_LODR":
+        try:
+            import kenlm
+        except ImportError:
+            print("Please install kenlm first. You can use")
+            print(" pip install https://github.com/kpu/kenlm/archive/master.zip")
+            print("to install it")
+            import sys
+
+            sys.exit(-1)
+        ngram_file_name = str(params.lang_dir / f"{params.tokens_ngram}gram.arpa")
+        logging.info(f"lm filename: {ngram_file_name}")
+        ngram_lm = kenlm.Model(ngram_file_name)
+        ngram_lm_scale = None  # use a list to search
+
+    elif params.decoding_method == "modified_beam_search_LODR":
+        lm_filename = f"{params.tokens_ngram}gram.fst.txt"
+        logging.info(f"Loading token level lm: {lm_filename}")
+        ngram_lm = NgramLm(
+            str(params.lang_dir / lm_filename),
+            backoff_id=params.backoff_id,
+            is_binary=False,
+        )
+        logging.info(f"num states: {ngram_lm.lm.num_states}")
+        ngram_lm_scale = params.ngram_lm_scale
+    else:
+        ngram_lm = None
+        ngram_lm_scale = None
+
+    if "fast_beam_search" in params.decoding_method:
+        if params.decoding_method == "fast_beam_search_nbest_LG":
+            lexicon = Lexicon(params.lang_dir)
+            word_table = lexicon.word_table
+            lg_filename = params.lang_dir / "LG.pt"
+            logging.info(f"Loading {lg_filename}")
+            decoding_graph = k2.Fsa.from_dict(
+                torch.load(lg_filename, map_location=device)
+            )
+            decoding_graph.scores *= params.ngram_lm_scale
+        else:
+            word_table = None
+            decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)
+    else:
+        decoding_graph = None
+        word_table = None
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    # we need cut ids to display recognition results.
+    args.return_cuts = True
+    ksponspeech = KsponSpeechAsrDataModule(args)
+
+    eval_clean_cuts = ksponspeech.eval_clean_cuts()
+    eval_other_cuts = ksponspeech.eval_other_cuts()
+
+    eval_clean_dl = ksponspeech.test_dataloaders(eval_clean_cuts)
+    eval_other_dl = ksponspeech.test_dataloaders(eval_other_cuts)
+
+    test_sets = ["eval_clean", "eval_other"]
+    test_dl = [eval_clean_dl, eval_other_dl]
+    import time
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        start = time.time()
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            sp=sp,
+            word_table=word_table,
+            decoding_graph=decoding_graph,
+            LM=LM,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+        )
+        logging.info(f"Elasped time for {test_set}: {time.time() - start}")
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    main()

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/decode_stream.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/decode_stream.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/decoder.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/decoder.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/encoder_interface.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/encoder_interface.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/export-onnx.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/export-onnx.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/export.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/export.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/joiner.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/joiner.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/model.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/model.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/onnx_check.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_check.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/onnx_model_wrapper.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_model_wrapper.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/onnx_pretrained.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/optim.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/optim.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/scaling.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/scaling.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/scaling_converter.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/scaling_converter.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/streaming_beam_search.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/streaming_beam_search.py

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/streaming_decode.py

@@ -0,0 +1,619 @@
+#!/usr/bin/env python3
+# Copyright    2024   (Author: SeungHyun Lee, Contacts: whsqkaak@naver.com)
+#
+# 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:
+./pruned_transducer_stateless7_streaming/streaming_decode.py \
+  --epoch 28 \
+  --avg 15 \
+  --decode-chunk-len 32 \
+  --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+  --decoding-method greedy_search \
+  --num-decode-streams 2000
+"""
+
+import argparse
+import logging
+import math
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import numpy as np
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import KsponSpeechAsrDataModule
+from decode_stream import DecodeStream
+from lhotse import CutSet, Fbank, FbankConfig
+from streaming_beam_search import (
+    fast_beam_search_one_best,
+    greedy_search,
+    modified_beam_search,
+)
+from torch.nn.utils.rnn import pad_sequence
+from train import add_model_arguments, get_params, get_transducer_model
+from zipformer import stack_states, unstack_states
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+LOG_EPS = math.log(1e-10)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=28,
+        help="""It specifies the checkpoint to use for decoding.
+        Note: Epoch counts from 1.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="pruned_transducer_stateless7_streaming/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Supported decoding methods are:
+        greedy_search
+        modified_beam_search
+        fast_beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "--num_active_paths",
+        type=int,
+        default=4,
+        help="""An interger indicating how many candidates we will keep for each
+        frame. Used only when --decoding-method is modified_beam_search.""",
+    )
+
+    parser.add_argument(
+        "--beam",
+        type=float,
+        default=4,
+        help="""A floating point value to calculate the cutoff score during beam
+        search (i.e., `cutoff = max-score - beam`), which is the same as the
+        `beam` in Kaldi.
+        Used only when --decoding-method is fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--max-contexts",
+        type=int,
+        default=4,
+        help="""Used only when --decoding-method is
+        fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--max-states",
+        type=int,
+        default=32,
+        help="""Used only when --decoding-method is
+        fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    parser.add_argument(
+        "--num-decode-streams",
+        type=int,
+        default=2000,
+        help="The number of streams that can be decoded parallel.",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def decode_one_chunk(
+    params: AttributeDict,
+    model: nn.Module,
+    decode_streams: List[DecodeStream],
+) -> List[int]:
+    """Decode one chunk frames of features for each decode_streams and
+    return the indexes of finished streams in a List.
+
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+      model:
+        The neural model.
+      decode_streams:
+        A List of DecodeStream, each belonging to a utterance.
+    Returns:
+      Return a List containing which DecodeStreams are finished.
+    """
+    device = model.device
+
+    features = []
+    feature_lens = []
+    states = []
+    processed_lens = []
+
+    for stream in decode_streams:
+        feat, feat_len = stream.get_feature_frames(params.decode_chunk_len)
+        features.append(feat)
+        feature_lens.append(feat_len)
+        states.append(stream.states)
+        processed_lens.append(stream.done_frames)
+
+    feature_lens = torch.tensor(feature_lens, device=device)
+    features = pad_sequence(features, batch_first=True, padding_value=LOG_EPS)
+
+    # We subsample features with ((x_len - 7) // 2 + 1) // 2 and the max downsampling
+    # factor in encoders is 8.
+    # After feature embedding (x_len - 7) // 2, we have (23 - 7) // 2 = 8.
+    tail_length = 23
+    if features.size(1) < tail_length:
+        pad_length = tail_length - features.size(1)
+        feature_lens += pad_length
+        features = torch.nn.functional.pad(
+            features,
+            (0, 0, 0, pad_length),
+            mode="constant",
+            value=LOG_EPS,
+        )
+
+    states = stack_states(states)
+    processed_lens = torch.tensor(processed_lens, device=device)
+
+    encoder_out, encoder_out_lens, new_states = model.encoder.streaming_forward(
+        x=features,
+        x_lens=feature_lens,
+        states=states,
+    )
+
+    encoder_out = model.joiner.encoder_proj(encoder_out)
+
+    if params.decoding_method == "greedy_search":
+        greedy_search(model=model, encoder_out=encoder_out, streams=decode_streams)
+    elif params.decoding_method == "fast_beam_search":
+        processed_lens = processed_lens + encoder_out_lens
+        fast_beam_search_one_best(
+            model=model,
+            encoder_out=encoder_out,
+            processed_lens=processed_lens,
+            streams=decode_streams,
+            beam=params.beam,
+            max_states=params.max_states,
+            max_contexts=params.max_contexts,
+        )
+    elif params.decoding_method == "modified_beam_search":
+        modified_beam_search(
+            model=model,
+            streams=decode_streams,
+            encoder_out=encoder_out,
+            num_active_paths=params.num_active_paths,
+        )
+    else:
+        raise ValueError(f"Unsupported decoding method: {params.decoding_method}")
+
+    states = unstack_states(new_states)
+
+    finished_streams = []
+    for i in range(len(decode_streams)):
+        decode_streams[i].states = states[i]
+        decode_streams[i].done_frames += encoder_out_lens[i]
+        if decode_streams[i].done:
+            finished_streams.append(i)
+
+    return finished_streams
+
+
+def decode_dataset(
+    cuts: CutSet,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    decoding_graph: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      cuts:
+        Lhotse Cutset containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      decoding_graph:
+        The decoding graph. Can be either a `k2.trivial_graph` or HLG, Used
+        only when --decoding_method is fast_beam_search.
+    Returns:
+      Return a dict, whose key may be "greedy_search" if greedy search
+      is used, or it may be "beam_7" if beam size of 7 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.
+    """
+    device = model.device
+
+    opts = FbankConfig(
+        device=device,
+        dither=0.0,
+        snip_edges=False,
+        sampling_rate=16000,
+        num_mel_bins=80,
+        high_freq=-400.0,
+    )
+
+    log_interval = 50
+
+    decode_results = []
+    # Contain decode streams currently running.
+    decode_streams = []
+    for num, cut in enumerate(cuts):
+        # each utterance has a DecodeStream.
+        initial_states = model.encoder.get_init_state(device=device)
+        decode_stream = DecodeStream(
+            params=params,
+            cut_id=cut.id,
+            initial_states=initial_states,
+            decoding_graph=decoding_graph,
+            device=device,
+        )
+
+        audio: np.ndarray = cut.load_audio()
+        # audio.shape: (1, num_samples)
+        assert len(audio.shape) == 2
+        assert audio.shape[0] == 1, "Should be single channel"
+        assert audio.dtype == np.float32, audio.dtype
+
+        # The trained model is using normalized samples
+        # - this is to avoid sending [-32k,+32k] signal in...
+        # - some lhotse AudioTransform classes can make the signal
+        #   be out of range [-1, 1], hence the tolerance 10
+        assert (
+            np.abs(audio).max() <= 10
+        ), "Should be normalized to [-1, 1], 10 for tolerance..."
+
+        samples = torch.from_numpy(audio).squeeze(0)
+
+        fbank = Fbank(opts)
+        feature = fbank.extract(samples.to(device), sampling_rate=16000)
+        decode_stream.set_features(feature, tail_pad_len=params.decode_chunk_len)
+        decode_stream.ground_truth = cut.supervisions[0].text
+
+        decode_streams.append(decode_stream)
+
+        while len(decode_streams) >= params.num_decode_streams:
+            finished_streams = decode_one_chunk(
+                params=params, model=model, decode_streams=decode_streams
+            )
+            for i in sorted(finished_streams, reverse=True):
+                decode_results.append(
+                    (
+                        decode_streams[i].id,
+                        decode_streams[i].ground_truth.split(),
+                        sp.decode(decode_streams[i].decoding_result()).split(),
+                    )
+                )
+                del decode_streams[i]
+
+        if num % log_interval == 0:
+            logging.info(f"Cuts processed until now is {num}.")
+
+    # decode final chunks of last sequences
+    while len(decode_streams):
+        finished_streams = decode_one_chunk(
+            params=params, model=model, decode_streams=decode_streams
+        )
+        for i in sorted(finished_streams, reverse=True):
+            decode_results.append(
+                (
+                    decode_streams[i].id,
+                    decode_streams[i].ground_truth.split(),
+                    sp.decode(decode_streams[i].decoding_result()).split(),
+                )
+            )
+            del decode_streams[i]
+
+    if params.decoding_method == "greedy_search":
+        key = "greedy_search"
+    elif params.decoding_method == "fast_beam_search":
+        key = (
+            f"beam_{params.beam}_"
+            f"max_contexts_{params.max_contexts}_"
+            f"max_states_{params.max_states}"
+        )
+    elif params.decoding_method == "modified_beam_search":
+        key = f"num_active_paths_{params.num_active_paths}"
+    else:
+        raise ValueError(f"Unsupported decoding method: {params.decoding_method}")
+    return {key: decode_results}
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[List[str], List[str]]]],
+):
+    test_set_cers = dict()
+    for key, results in results_dict.items():
+        recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out CERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
+        with open(errs_filename, "w") as f:
+            cer = write_error_stats(
+                f,
+                f"{test_set_name}-{key}",
+                results,
+                enable_log=True,
+                compute_CER=True,
+            )
+            test_set_cers[key] = cer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_cers = sorted(test_set_cers.items(), key=lambda x: x[1])
+    errs_info = params.res_dir / f"cer-summary-{test_set_name}-{params.suffix}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tCER", file=f)
+        for key, val in test_set_cers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, CER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_cers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    KsponSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    params.res_dir = params.exp_dir / "streaming" / params.decoding_method
+
+    if params.iter > 0:
+        params.suffix = f"iter-{params.iter}-avg-{params.avg}"
+    else:
+        params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+
+    # for streaming
+    params.suffix += f"-streaming-chunk-size-{params.decode_chunk_len}"
+
+    # for fast_beam_search
+    if params.decoding_method == "fast_beam_search":
+        params.suffix += f"-beam-{params.beam}"
+        params.suffix += f"-max-contexts-{params.max_contexts}"
+        params.suffix += f"-max-states-{params.max_states}"
+
+    if params.use_averaged_model:
+        params.suffix += "-use-averaged-model"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <unk> is defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.unk_id = sp.piece_to_id("<unk>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif 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.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.to(device)
+    model.eval()
+    model.device = device
+
+    decoding_graph = None
+    if params.decoding_method == "fast_beam_search":
+        decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    ksponspeech = KsponSpeechAsrDataModule(args)
+
+    eval_clean_cuts = ksponspeech.eval_clean_cuts()
+    eval_other_cuts = ksponspeech.eval_other_cuts()
+
+    test_sets = ["eval_clean", "eval_other"]
+    test_cuts = [eval_clean_cuts, eval_other_cuts]
+
+    for test_set, test_cut in zip(test_sets, test_cuts):
+        results_dict = decode_dataset(
+            cuts=test_cut,
+            params=params,
+            model=model,
+            sp=sp,
+            decoding_graph=decoding_graph,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    main()

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/test_model.py

@@ -0,0 +1,187 @@
+#!/usr/bin/env python3
+# Copyright    2024   (Author: SeungHyun Lee, Contacts: whsqkaak@naver.com)
+#
+# 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.
+
+
+"""
+To run this file, do:
+
+    cd icefall/egs/ksponspeech/ASR
+    python ./pruned_transducer_stateless7_streaming/test_model.py
+"""
+
+import torch
+from scaling_converter import convert_scaled_to_non_scaled
+from train import get_params, get_transducer_model
+
+
+def test_model():
+    params = get_params()
+    params.vocab_size = 500
+    params.blank_id = 0
+    params.context_size = 2
+    params.num_encoder_layers = "2,4,3,2,4"
+    params.feedforward_dims = "1024,1024,2048,2048,1024"
+    params.nhead = "8,8,8,8,8"
+    params.encoder_dims = "384,384,384,384,384"
+    params.attention_dims = "192,192,192,192,192"
+    params.encoder_unmasked_dims = "256,256,256,256,256"
+    params.zipformer_downsampling_factors = "1,2,4,8,2"
+    params.cnn_module_kernels = "31,31,31,31,31"
+    params.decoder_dim = 512
+    params.joiner_dim = 512
+    params.num_left_chunks = 4
+    params.short_chunk_size = 50
+    params.decode_chunk_len = 32
+    model = get_transducer_model(params)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    print(f"Number of model parameters: {num_param}")
+
+    # Test jit script
+    convert_scaled_to_non_scaled(model, inplace=True)
+    # We won't use the forward() method of the model in C++, so just ignore
+    # it here.
+    # Otherwise, one of its arguments is a ragged tensor and is not
+    # torch scriptabe.
+    model.__class__.forward = torch.jit.ignore(model.__class__.forward)
+    print("Using torch.jit.script")
+    model = torch.jit.script(model)
+
+
+def test_model_small():
+    params = get_params()
+    params.vocab_size = 500
+    params.blank_id = 0
+    params.context_size = 2
+    params.num_encoder_layers = "2,2,2,2,2"
+    params.feedforward_dims = "256,256,512,512,256"
+    params.nhead = "4,4,4,4,4"
+    params.encoder_dims = "128,128,128,128,128"
+    params.attention_dims = "96,96,96,96,96"
+    params.encoder_unmasked_dims = "96,96,96,96,96"
+    params.zipformer_downsampling_factors = "1,2,4,8,2"
+    params.cnn_module_kernels = "31,31,31,31,31"
+    params.decoder_dim = 320
+    params.joiner_dim = 320
+    params.num_left_chunks = 4
+    params.short_chunk_size = 50
+    params.decode_chunk_len = 32
+    model = get_transducer_model(params)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    print(f"Number of model parameters: {num_param}")
+    import pdb
+
+    pdb.set_trace()
+
+    # Test jit script
+    convert_scaled_to_non_scaled(model, inplace=True)
+    # We won't use the forward() method of the model in C++, so just ignore
+    # it here.
+    # Otherwise, one of its arguments is a ragged tensor and is not
+    # torch scriptabe.
+    model.__class__.forward = torch.jit.ignore(model.__class__.forward)
+    print("Using torch.jit.script")
+    model = torch.jit.script(model)
+
+
+def test_model_jit_trace():
+    params = get_params()
+    params.vocab_size = 500
+    params.blank_id = 0
+    params.context_size = 2
+    params.num_encoder_layers = "2,4,3,2,4"
+    params.feedforward_dims = "1024,1024,2048,2048,1024"
+    params.nhead = "8,8,8,8,8"
+    params.encoder_dims = "384,384,384,384,384"
+    params.attention_dims = "192,192,192,192,192"
+    params.encoder_unmasked_dims = "256,256,256,256,256"
+    params.zipformer_downsampling_factors = "1,2,4,8,2"
+    params.cnn_module_kernels = "31,31,31,31,31"
+    params.decoder_dim = 512
+    params.joiner_dim = 512
+    params.num_left_chunks = 4
+    params.short_chunk_size = 50
+    params.decode_chunk_len = 32
+    model = get_transducer_model(params)
+    model.eval()
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    print(f"Number of model parameters: {num_param}")
+
+    convert_scaled_to_non_scaled(model, inplace=True)
+
+    # Test encoder
+    def _test_encoder():
+        encoder = model.encoder
+        assert encoder.decode_chunk_size == params.decode_chunk_len // 2, (
+            encoder.decode_chunk_size,
+            params.decode_chunk_len,
+        )
+        T = params.decode_chunk_len + 7
+
+        x = torch.zeros(1, T, 80, dtype=torch.float32)
+        x_lens = torch.full((1,), T, dtype=torch.int32)
+        states = encoder.get_init_state(device=x.device)
+        encoder.__class__.forward = encoder.__class__.streaming_forward
+        traced_encoder = torch.jit.trace(encoder, (x, x_lens, states))
+
+        states1 = encoder.get_init_state(device=x.device)
+        states2 = traced_encoder.get_init_state(device=x.device)
+        for i in range(5):
+            x = torch.randn(1, T, 80, dtype=torch.float32)
+            x_lens = torch.full((1,), T, dtype=torch.int32)
+            y1, _, states1 = encoder.streaming_forward(x, x_lens, states1)
+            y2, _, states2 = traced_encoder(x, x_lens, states2)
+            assert torch.allclose(y1, y2, atol=1e-6), (i, (y1 - y2).abs().mean())
+
+    # Test decoder
+    def _test_decoder():
+        decoder = model.decoder
+        y = torch.zeros(10, decoder.context_size, dtype=torch.int64)
+        need_pad = torch.tensor([False])
+
+        traced_decoder = torch.jit.trace(decoder, (y, need_pad))
+        d1 = decoder(y, need_pad)
+        d2 = traced_decoder(y, need_pad)
+        assert torch.equal(d1, d2), (d1 - d2).abs().mean()
+
+    # Test joiner
+    def _test_joiner():
+        joiner = model.joiner
+        encoder_out_dim = joiner.encoder_proj.weight.shape[1]
+        decoder_out_dim = joiner.decoder_proj.weight.shape[1]
+        encoder_out = torch.rand(1, encoder_out_dim, dtype=torch.float32)
+        decoder_out = torch.rand(1, decoder_out_dim, dtype=torch.float32)
+
+        traced_joiner = torch.jit.trace(joiner, (encoder_out, decoder_out))
+        j1 = joiner(encoder_out, decoder_out)
+        j2 = traced_joiner(encoder_out, decoder_out)
+        assert torch.equal(j1, j2), (j1 - j2).abs().mean()
+
+    _test_encoder()
+    _test_decoder()
+    _test_joiner()
+
+
+def main():
+    test_model_small()
+    test_model_jit_trace()
+
+
+if __name__ == "__main__":
+    main()

egs/ksponspeech/ASR/pruned_transducer_stateless7_streaming/zipformer.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless7_streaming/zipformer.py

egs/ksponspeech/ASR/shared

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