Merge branch 'k2-fsa:master' into tiny

egs/aishell/ASR/RESULTS.md

@@ -15,6 +15,8 @@ It uses pruned RNN-T.
 |------------------------|------|------|---------------------------------------|
 | greedy search          | 5.39 | 5.09 | --epoch 29 --avg 5 --max-duration 600 |
 | modified beam search   | 5.05 | 4.79 | --epoch 29 --avg 5 --max-duration 600 |
+| modified beam search + RNNLM shallow fusion   | 4.73 | 4.53 | --epoch 29 --avg 5 --max-duration 600 |
+| modified beam search + LODR   | 4.57 | 4.37 | --epoch 29 --avg 5 --max-duration 600 |
 | fast beam search       | 5.13 | 4.91 | --epoch 29 --avg 5 --max-duration 600 |
 
 Training command is:
@@ -73,6 +75,78 @@ for epoch in 29; do
 done
 ```
 
+We provide the option of shallow fusion with a RNN language model. The pre-trained language model is 
+available at <https://huggingface.co/marcoyang/icefall-aishell-rnn-lm>. To decode with the language model,
+please use the following command:
+
+```bash
+# download pre-trained model
+git lfs install
+git clone https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
+
+aishell_exp=icefall-aishell-pruned-transducer-stateless3-2022-06-20/
+
+pushd ${aishell_exp}/exp
+ln -s pretrained-epoch-29-avg-5-torch-1.10.0.pt epoch-99.pt
+popd
+
+# download RNN LM
+git lfs install
+git clone https://huggingface.co/marcoyang/icefall-aishell-rnn-lm
+rnnlm_dir=icefall-aishell-rnn-lm
+
+# RNNLM shallow fusion
+for lm_scale in $(seq 0.26 0.02 0.34); do
+  python ./pruned_transducer_stateless3/decode.py \
+      --epoch 99 \
+      --avg 1 \
+      --lang-dir ${aishell_exp}/data/lang_char \
+      --exp-dir ${aishell_exp}/exp \
+      --use-averaged-model False \
+      --decoding-method modified_beam_search_lm_shallow_fusion \
+      --use-shallow-fusion 1 \
+      --lm-type rnn \
+      --lm-exp-dir ${rnnlm_dir}/exp \
+      --lm-epoch 99 \
+      --lm-scale $lm_scale \
+      --lm-avg 1 \
+      --rnn-lm-embedding-dim 2048 \
+      --rnn-lm-hidden-dim 2048 \
+      --rnn-lm-num-layers 2 \
+      --lm-vocab-size 4336
+done
+
+# RNNLM Low-order density ratio (LODR) with a 2-gram
+
+cp ${rnnlm_dir}/2gram.fst.txt ${aishell_exp}/data/lang_char/2gram.fst.txt
+
+for lm_scale in 0.48; do
+  for LODR_scale in -0.28; do
+    python ./pruned_transducer_stateless3/decode.py \
+        --epoch 99 \
+        --avg 1 \
+        --lang-dir ${aishell_exp}/data/lang_char \
+        --exp-dir ${aishell_exp}/exp \
+        --use-averaged-model False \
+        --decoding-method modified_beam_search_LODR \
+        --use-shallow-fusion 1 \
+        --lm-type rnn \
+        --lm-exp-dir ${rnnlm_dir}/exp \
+        --lm-epoch 99 \
+        --lm-scale $lm_scale \
+        --lm-avg 1 \
+        --rnn-lm-embedding-dim 2048 \
+        --rnn-lm-hidden-dim 2048 \
+        --rnn-lm-num-layers 2 \
+        --lm-vocab-size 4336 \
+        --tokens-ngram 2 \
+        --backoff-id 4336 \
+        --ngram-lm-scale $LODR_scale
+  done
+done
+
+```
+
 Pretrained models, training logs, decoding logs, and decoding results
 are available at
 <https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20>

egs/aishell/ASR/local/sort_lm_training_data.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/local/sort_lm_training_data.py

egs/aishell/ASR/prepare.sh

@@ -231,11 +231,13 @@ if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
     cp $lang_phone_dir/transcript_words.txt $dl_dir/lm/aishell-train-word.txt
   fi
   
+  # training words
   ./local/prepare_char_lm_training_data.py \
     --lang-char data/lang_char \
     --lm-data $dl_dir/lm/aishell-train-word.txt \
     --lm-archive $out_dir/lm_data.pt
 
+  # valid words
   if [ ! -f $dl_dir/lm/aishell-valid-word.txt ]; then
     aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
     aishell_valid_uid=$dl_dir/aishell/data_aishell/transcript/aishell_valid_uid
@@ -249,6 +251,7 @@ if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
     --lm-data $dl_dir/lm/aishell-valid-word.txt \
     --lm-archive $out_dir/lm_data_valid.pt
 
+  # test words
   if [ ! -f $dl_dir/lm/aishell-test-word.txt ]; then
     aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
     aishell_test_uid=$dl_dir/aishell/data_aishell/transcript/aishell_test_uid
@@ -303,9 +306,9 @@ if [ $stage -le 11 ] && [ $stop_stage -ge 11 ]; then
     --hidden-dim 512 \
     --num-layers 2 \
     --batch-size 400 \
-    --exp-dir rnnlm_char/exp \
+    --exp-dir rnnlm_char/exp_aishell1_small \
-    --lm-data data/lm_training_char/sorted_lm_data.pt \
+    --lm-data data/lm_char/sorted_lm_data_aishell1.pt \
-    --lm-data-valid data/lm_training_char/sorted_lm_data-valid.pt \
+    --lm-data-valid data/lm_char/sorted_lm_data_valid.pt \
     --vocab-size 4336 \
     --master-port 12345
 fi

egs/aishell/ASR/pruned_transducer_stateless3/decode.py

@@ -54,6 +54,40 @@ Usage:
     --beam 4 \
     --max-contexts 4 \
     --max-states 8
+
+(5) modified beam search (with LM shallow fusion)
+./pruned_transducer_stateless3/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless3/exp \
+    --max-duration 600 \
+    --decoding-method modified_beam_search_lm_shallow_fusion \
+    --beam-size 4 \
+    --lm-type rnn \
+    --lm-scale 0.3 \
+    --lm-exp-dir /path/to/LM \
+    --rnn-lm-epoch 99 \
+    --rnn-lm-avg 1 \
+    --rnn-lm-num-layers 3 \
+    --rnn-lm-tie-weights 1
+
+(6) modified beam search with LM shallow fusion + LODR
+./pruned_transducer_stateless3/decode.py \
+    --epoch 28 \
+    --avg 15 \
+    --max-duration 600 \
+    --exp-dir ./pruned_transducer_stateless3/exp \
+    --decoding-method modified_beam_search_LODR \
+    --beam-size 4 \
+    --lm-type rnn \
+    --lm-scale 0.48 \
+    --lm-exp-dir /path/to/LM \
+    --rnn-lm-epoch 99 \
+    --rnn-lm-avg 1 \
+    --rnn-lm-num-layers 3 \
+    --rnn-lm-tie-weights 1
+    --tokens-ngram 2 \
+    --ngram-lm-scale -0.28 \
 """
 
 
@@ -74,9 +108,12 @@ from beam_search import (
     greedy_search,
     greedy_search_batch,
     modified_beam_search,
+    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,
@@ -212,6 +249,60 @@ def get_parser():
         Used only when --decoding_method is greedy_search""",
     )
 
+    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="""Token Ngram used for rescoring.
+            Used only when the decoding method is
+            modified_beam_search_ngram_rescoring""",
+    )
+
+    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(
+        "--backoff-id",
+        type=int,
+        default=500,
+        help="""ID of the backoff symbol.
+                Used only when the decoding method is
+                modified_beam_search_ngram_rescoring""",
+    )
+
     add_model_arguments(parser)
 
     return parser
@@ -223,6 +314,9 @@ def decode_one_batch(
     token_table: k2.SymbolTable,
     batch: dict,
     decoding_graph: Optional[k2.Fsa] = None,
+    ngram_lm: Optional[NgramLm] = None,
+    ngram_lm_scale: float = 1.0,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[List[str]]]:
     """Decode one batch and return the result in a dict. The dict has the
     following format:
@@ -287,6 +381,24 @@ def decode_one_batch(
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
         )
+    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,
+        )
+    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,
+        )
     else:
         hyp_tokens = []
         batch_size = encoder_out.size(0)
@@ -334,6 +446,9 @@ def decode_dataset(
     model: nn.Module,
     token_table: k2.SymbolTable,
     decoding_graph: Optional[k2.Fsa] = None,
+    ngram_lm: Optional[NgramLm] = None,
+    ngram_lm_scale: float = 1.0,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
     """Decode dataset.
 
@@ -379,6 +494,9 @@ def decode_dataset(
             token_table=token_table,
             decoding_graph=decoding_graph,
             batch=batch,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+            LM=LM,
         )
 
         for name, hyps in hyps_dict.items():
@@ -445,6 +563,7 @@ def save_results(
 def main():
     parser = get_parser()
     AsrDataModule.add_arguments(parser)
+    LmScorer.add_arguments(parser)
     args = parser.parse_args()
     args.exp_dir = Path(args.exp_dir)
     args.lang_dir = Path(args.lang_dir)
@@ -458,6 +577,8 @@ def main():
         "beam_search",
         "fast_beam_search",
         "modified_beam_search",
+        "modified_beam_search_LODR",
+        "modified_beam_search_lm_shallow_fusion",
     )
     params.res_dir = params.exp_dir / params.decoding_method
 
@@ -479,6 +600,19 @@ def main():
     if params.use_averaged_model:
         params.suffix += "-use-averaged-model"
 
+    if "ngram" in params.decoding_method:
+        params.suffix += f"-ngram-lm-scale-{params.ngram_lm_scale}"
+    if params.use_shallow_fusion:
+        if params.lm_type == "rnn":
+            params.suffix += f"-rnnlm-lm-scale-{params.lm_scale}"
+        elif params.lm_type == "transformer":
+            params.suffix += f"-transformer-lm-scale-{params.lm_scale}"
+
+        if "LODR" in params.decoding_method:
+            params.suffix += (
+                f"-LODR-{params.tokens_ngram}gram-scale-{params.ngram_lm_scale}"
+            )
+
     setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
     logging.info("Decoding started")
 
@@ -588,6 +722,35 @@ def main():
     else:
         decoding_graph = None
 
+    # only load N-gram LM when needed
+    if "ngram" in params.decoding_method or "LODR" in params.decoding_method:
+        lm_filename = params.lang_dir / f"{params.tokens_ngram}gram.fst.txt"
+        logging.info(f"lm filename: {lm_filename}")
+        ngram_lm = NgramLm(
+            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
+
+    # only load the neural network LM if doing shallow fusion
+    if params.use_shallow_fusion:
+        LM = LmScorer(
+            lm_type=params.lm_type,
+            params=params,
+            device=device,
+            lm_scale=params.lm_scale,
+        )
+        LM.to(device)
+        LM.eval()
+
+    else:
+        LM = None
+
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
@@ -610,6 +773,9 @@ def main():
             model=model,
             token_table=lexicon.token_table,
             decoding_graph=decoding_graph,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+            LM=LM,
         )
 
         save_results(

egs/aishell/ASR/pruned_transducer_stateless3/exp-context-size-1

@@ -1 +0,0 @@
-/ceph-fj/fangjun/open-source/icefall-aishell/egs/aishell/ASR/pruned_transducer_stateless3/exp-context-size-1

egs/aishell/ASR/pruned_transducer_stateless3/export.py

@@ -48,6 +48,7 @@ import logging
 from pathlib import Path
 
 import torch
+from scaling_converter import convert_scaled_to_non_scaled
 from train import add_model_arguments, get_params, get_transducer_model
 
 from icefall.checkpoint import (
@@ -244,6 +245,7 @@ def main():
     model.eval()
 
     if params.jit:
+        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

egs/aishell/ASR/pruned_transducer_stateless3/lstmp.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless3/lstmp.py

egs/aishell/ASR/pruned_transducer_stateless3/scaling_converter.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless3/scaling_converter.py

egs/commonvoice/ASR/local/compute_fbank_commonvoice_dev_test.py

@@ -0,0 +1,107 @@
+#!/usr/bin/env python3
+# Copyright    2023  Xiaomi Corp.        (authors: Yifan Yang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This file computes fbank features of the CommonVoice dataset.
+It looks for manifests in the directory data/${lang}/manifests.
+
+The generated fbank features are saved in data/${lang}/fbank.
+"""
+
+import argparse
+import logging
+import os
+from pathlib import Path
+from typing import Optional
+
+import torch
+from filter_cuts import filter_cuts
+from lhotse import CutSet, KaldifeatFbank, KaldifeatFbankConfig, LilcomChunkyWriter
+
+# 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(
+        "--language",
+        type=str,
+        help="""Language of Common Voice""",
+    )
+
+    return parser.parse_args()
+
+
+def compute_fbank_commonvoice_dev_test(language: str):
+    src_dir = Path(f"data/{language}/manifests")
+    output_dir = Path(f"data/{language}/fbank")
+    num_workers = 42
+    batch_duration = 600
+
+    subsets = ("dev", "test")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+    extractor = KaldifeatFbank(KaldifeatFbankConfig(device=device))
+
+    logging.info(f"device: {device}")
+
+    for partition in subsets:
+        cuts_path = output_dir / f"cv-{language}_cuts_{partition}.jsonl.gz"
+        if cuts_path.is_file():
+            logging.info(f"{partition} already exists - skipping.")
+            continue
+
+        raw_cuts_path = output_dir / f"cv-{language}_cuts_{partition}_raw.jsonl.gz"
+
+        logging.info(f"Loading {raw_cuts_path}")
+        cut_set = CutSet.from_file(raw_cuts_path)
+
+        logging.info("Splitting cuts into smaller chunks")
+        cut_set = cut_set.trim_to_supervisions(
+            keep_overlapping=False, min_duration=None
+        )
+
+        logging.info("Computing features")
+        cut_set = cut_set.compute_and_store_features_batch(
+            extractor=extractor,
+            storage_path=f"{output_dir}/cv-{language}_feats_{partition}",
+            num_workers=num_workers,
+            batch_duration=batch_duration,
+            storage_type=LilcomChunkyWriter,
+            overwrite=True,
+        )
+
+        logging.info(f"Saving to {cuts_path}")
+        cut_set.to_file(cuts_path)
+
+
+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_commonvoice_dev_test(language=args.language)

egs/commonvoice/ASR/local/compute_fbank_commonvoice_splits.py

@@ -0,0 +1,157 @@
+#!/usr/bin/env python3
+# Copyright    2023  Xiaomi Corp.             (Yifan Yang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import logging
+from datetime import datetime
+from pathlib import Path
+
+import torch
+from lhotse import (
+    CutSet,
+    KaldifeatFbank,
+    KaldifeatFbankConfig,
+    LilcomChunkyWriter,
+    set_audio_duration_mismatch_tolerance,
+    set_caching_enabled,
+)
+
+# 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(
+        "--language",
+        type=str,
+        help="""Language of Common Voice""",
+    )
+
+    parser.add_argument(
+        "--num-workers",
+        type=int,
+        default=20,
+        help="Number of dataloading workers used for reading the audio.",
+    )
+
+    parser.add_argument(
+        "--batch-duration",
+        type=float,
+        default=600.0,
+        help="The maximum number of audio seconds in a batch."
+        "Determines batch size dynamically.",
+    )
+
+    parser.add_argument(
+        "--num-splits",
+        type=int,
+        required=True,
+        help="The number of splits of the train subset",
+    )
+
+    parser.add_argument(
+        "--start",
+        type=int,
+        default=0,
+        help="Process pieces starting from this number (inclusive).",
+    )
+
+    parser.add_argument(
+        "--stop",
+        type=int,
+        default=-1,
+        help="Stop processing pieces until this number (exclusive).",
+    )
+
+    return parser.parse_args()
+
+
+def compute_fbank_commonvoice_splits(args):
+    subset = "train"
+    num_splits = args.num_splits
+    language = args.language
+    output_dir = f"data/{language}/fbank/{subset}_split_{num_splits}"
+    output_dir = Path(output_dir)
+    assert output_dir.exists(), f"{output_dir} does not exist!"
+
+    num_digits = len(str(num_splits))
+
+    start = args.start
+    stop = args.stop
+    if stop < start:
+        stop = num_splits
+
+    stop = min(stop, num_splits)
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+    extractor = KaldifeatFbank(KaldifeatFbankConfig(device=device))
+    logging.info(f"device: {device}")
+
+    set_audio_duration_mismatch_tolerance(0.01)  # 10ms tolerance
+    set_caching_enabled(False)
+    for i in range(start, stop):
+        idx = f"{i + 1}".zfill(num_digits)
+        logging.info(f"Processing {idx}/{num_splits}")
+
+        cuts_path = output_dir / f"cv-{language}_cuts_{subset}.{idx}.jsonl.gz"
+        if cuts_path.is_file():
+            logging.info(f"{cuts_path} exists - skipping")
+            continue
+
+        raw_cuts_path = output_dir / f"cv-{language}_cuts_{subset}_raw.{idx}.jsonl.gz"
+
+        logging.info(f"Loading {raw_cuts_path}")
+        cut_set = CutSet.from_file(raw_cuts_path)
+
+        logging.info("Splitting cuts into smaller chunks.")
+        cut_set = cut_set.trim_to_supervisions(
+            keep_overlapping=False, min_duration=None
+        )
+
+        logging.info("Computing features")
+        cut_set = cut_set.compute_and_store_features_batch(
+            extractor=extractor,
+            storage_path=f"{output_dir}/cv-{language}_feats_{subset}_{idx}",
+            num_workers=args.num_workers,
+            batch_duration=args.batch_duration,
+            storage_type=LilcomChunkyWriter,
+            overwrite=True,
+        )
+
+        logging.info(f"Saving to {cuts_path}")
+        cut_set.to_file(cuts_path)
+
+
+def 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_commonvoice_splits(args)
+
+
+if __name__ == "__main__":
+    main()

egs/commonvoice/ASR/local/compute_fbank_musan.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/local/compute_fbank_musan.py

egs/commonvoice/ASR/local/filter_cuts.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/local/filter_cuts.py

egs/commonvoice/ASR/local/preprocess_commonvoice.py

@@ -0,0 +1,119 @@
+#!/usr/bin/env python3
+# Copyright    2023  Xiaomi Corp.        (authors: Yifan Yang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import logging
+from pathlib import Path
+from typing import Optional
+
+from lhotse import CutSet, SupervisionSegment
+from lhotse.recipes.utils import read_manifests_if_cached
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        help="""Dataset parts to compute fbank. If None, we will use all""",
+    )
+
+    parser.add_argument(
+        "--language",
+        type=str,
+        help="""Language of Common Voice""",
+    )
+
+    return parser.parse_args()
+
+
+def preprocess_commonvoice(
+    language: str,
+    dataset: Optional[str] = None,
+):
+    src_dir = Path(f"data/{language}/manifests")
+    output_dir = Path(f"data/{language}/fbank")
+    output_dir.mkdir(exist_ok=True)
+
+    if dataset is None:
+        dataset_parts = (
+            "dev",
+            "test",
+            "train",
+        )
+    else:
+        dataset_parts = dataset.split(" ", -1)
+
+    logging.info("Loading manifest")
+    prefix = f"cv-{language}"
+    suffix = "jsonl.gz"
+    manifests = read_manifests_if_cached(
+        dataset_parts=dataset_parts,
+        output_dir=src_dir,
+        suffix=suffix,
+        prefix=prefix,
+    )
+    assert manifests is not None
+
+    assert len(manifests) == len(dataset_parts), (
+        len(manifests),
+        len(dataset_parts),
+        list(manifests.keys()),
+        dataset_parts,
+    )
+
+    for partition, m in manifests.items():
+        logging.info(f"Processing {partition}")
+        raw_cuts_path = output_dir / f"{prefix}_cuts_{partition}_raw.{suffix}"
+        if raw_cuts_path.is_file():
+            logging.info(f"{partition} already exists - skipping")
+            continue
+
+        # Create long-recording cut manifests.
+        cut_set = CutSet.from_manifests(
+            recordings=m["recordings"],
+            supervisions=m["supervisions"],
+        ).resample(16000)
+
+        # Run data augmentation that needs to be done in the
+        # time domain.
+        if "train" in partition:
+            logging.info(
+                f"Speed perturb for {partition} with factors 0.9 and 1.1 "
+                "(Perturbing may take 2 minutes and saving may take 7 minutes)"
+            )
+            cut_set = cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
+        logging.info(f"Saving to {raw_cuts_path}")
+        cut_set.to_file(raw_cuts_path)
+
+
+def 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))
+    preprocess_commonvoice(
+        language=args.language,
+        dataset=args.dataset,
+    )
+    logging.info("Done")
+
+
+if __name__ == "__main__":
+    main()

egs/commonvoice/ASR/prepare.sh

@@ -0,0 +1,156 @@
+#!/usr/bin/env bash
+
+set -eou pipefail
+
+nj=16
+stage=-1
+stop_stage=100
+
+# Split data/${lang}set to this number of pieces
+# This is to avoid OOM during feature extraction.
+num_splits=1000
+
+# We assume dl_dir (download dir) contains the following
+# directories and files. If not, they will be downloaded
+# by this script automatically.
+#
+#  - $dl_dir/$release/$lang
+#      This directory contains the following files downloaded from
+#       https://mozilla-common-voice-datasets.s3.dualstack.us-west-2.amazonaws.com/${release}/${release}-${lang}.tar.gz
+#
+#     - clips
+#     - dev.tsv
+#     - invalidated.tsv
+#     - other.tsv
+#     - reported.tsv
+#     - test.tsv
+#     - train.tsv
+#     - validated.tsv
+#
+#  - $dl_dir/musan
+#      This directory contains the following directories downloaded from
+#       http://www.openslr.org/17/
+#
+#     - music
+#     - noise
+#     - speech
+
+dl_dir=$PWD/download
+release=cv-corpus-13.0-2023-03-09
+lang=en
+
+. shared/parse_options.sh || exit 1
+
+# vocab size for sentence piece models.
+# It will generate data/${lang}/lang_bpe_xxx,
+# data/${lang}/lang_bpe_yyy if the array contains xxx, yyy
+vocab_sizes=(
+  # 5000
+  # 2000
+  # 1000
+  500
+)
+
+# All files generated by this script are saved in "data/${lang}".
+# You can safely remove "data/${lang}" and rerun this script to regenerate it.
+mkdir -p data/${lang}
+
+log() {
+  # This function is from espnet
+  local fname=${BASH_SOURCE[1]##*/}
+  echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+log "dl_dir: $dl_dir"
+
+if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
+  log "Stage 0: Download data"
+
+  # If you have pre-downloaded it to /path/to/$release,
+  # you can create a symlink
+  #
+  #   ln -sfv /path/to/$release $dl_dir/$release
+  #
+  if [ ! -d $dl_dir/$release/$lang/clips ]; then
+    lhotse download commonvoice --languages $lang --release $release $dl_dir
+  fi
+
+  # If you have pre-downloaded it to /path/to/musan,
+  # you can create a symlink
+  #
+  #   ln -sfv /path/to/musan $dl_dir/
+  #
+  if [ ! -d $dl_dir/musan ]; then
+    lhotse download musan $dl_dir
+  fi
+fi
+
+if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
+  log "Stage 1: Prepare CommonVoice manifest"
+  # We assume that you have downloaded the CommonVoice corpus
+  # to $dl_dir/$release
+  mkdir -p data/${lang}/manifests
+  if [ ! -e data/${lang}/manifests/.cv-${lang}.done ]; then
+    lhotse prepare commonvoice --language $lang -j $nj $dl_dir/$release data/${lang}/manifests
+    touch data/${lang}/manifests/.cv-${lang}.done
+  fi
+fi
+
+if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
+  log "Stage 2: Prepare musan manifest"
+  # We assume that you have downloaded the musan corpus
+  # to data/musan
+  mkdir -p data/manifests
+  if [ ! -e data/manifests/.musan.done ]; then
+    lhotse prepare musan $dl_dir/musan data/manifests
+    touch data/manifests/.musan.done
+  fi
+fi
+
+if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
+  log "Stage 3: Preprocess CommonVoice manifest"
+  if [ ! -e data/${lang}/fbank/.preprocess_complete ]; then
+    ./local/preprocess_commonvoice.py  --language $lang
+    touch data/${lang}/fbank/.preprocess_complete
+  fi
+fi
+
+if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
+  log "Stage 4: Compute fbank for dev and test subsets of CommonVoice"
+  mkdir -p data/${lang}/fbank
+  if [ ! -e data/${lang}/fbank/.cv-${lang}_dev_test.done ]; then
+    ./local/compute_fbank_commonvoice_dev_test.py --language $lang
+    touch data/${lang}/fbank/.cv-${lang}_dev_test.done
+  fi
+fi
+
+if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
+  log "Stage 5: Split train subset into ${num_splits} pieces"
+  split_dir=data/${lang}/fbank/train_split_${num_splits}
+  if [ ! -e $split_dir/.cv-${lang}_train_split.done ]; then
+    lhotse split $num_splits ./data/${lang}/fbank/cv-${lang}_cuts_train_raw.jsonl.gz $split_dir
+    touch $split_dir/.cv-${lang}_train_split.done
+  fi
+fi
+
+if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
+  log "Stage 6: Compute features for train subset of CommonVoice"
+  if [ ! -e data/${lang}/fbank/.cv-${lang}_train.done ]; then
+    ./local/compute_fbank_commonvoice_splits.py \
+      --num-workers $nj \
+      --batch-duration 600 \
+      --start 0 \
+      --num-splits $num_splits \
+      --language $lang
+    touch data/${lang}/fbank/.cv-${lang}_train.done
+  fi
+fi
+
+if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
+  log "Stage 7: Compute fbank for musan"
+  mkdir -p data/fbank
+  if [ ! -e data/fbank/.musan.done ]; then
+    ./local/compute_fbank_musan.py
+    touch data/fbank/.musan.done
+  fi
+fi

egs/commonvoice/ASR/shared

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

egs/gigaspeech/ASR/local/generate_unique_lexicon.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/local/generate_unique_lexicon.py

egs/gigaspeech/ASR/pruned_transducer_stateless2/decode.py

@@ -19,40 +19,40 @@
 Usage:
 (1) greedy search
 ./pruned_transducer_stateless2/decode.py \
-        --epoch 28 \
+    --epoch 28 \
-        --avg 15 \
+    --avg 15 \
-        --exp-dir ./pruned_transducer_stateless2/exp \
+    --exp-dir ./pruned_transducer_stateless2/exp \
-        --max-duration 600 \
+    --max-duration 600 \
-        --decoding-method greedy_search
+    --decoding-method greedy_search
 
 (2) beam search
 ./pruned_transducer_stateless2/decode.py \
-        --epoch 28 \
+    --epoch 28 \
-        --avg 15 \
+    --avg 15 \
-        --exp-dir ./pruned_transducer_stateless2/exp \
+    --exp-dir ./pruned_transducer_stateless2/exp \
-        --max-duration 600 \
+    --max-duration 600 \
-        --decoding-method beam_search \
+    --decoding-method beam_search \
-        --beam-size 4
+    --beam-size 4
 
 (3) modified beam search
 ./pruned_transducer_stateless2/decode.py \
-        --epoch 28 \
+    --epoch 28 \
-        --avg 15 \
+    --avg 15 \
-        --exp-dir ./pruned_transducer_stateless2/exp \
+    --exp-dir ./pruned_transducer_stateless2/exp \
-        --max-duration 600 \
+    --max-duration 600 \
-        --decoding-method modified_beam_search \
+    --decoding-method modified_beam_search \
-        --beam-size 4
+    --beam-size 4
 
 (4) fast beam search
 ./pruned_transducer_stateless2/decode.py \
-        --epoch 28 \
+    --epoch 28 \
-        --avg 15 \
+    --avg 15 \
-        --exp-dir ./pruned_transducer_stateless2/exp \
+    --exp-dir ./pruned_transducer_stateless2/exp \
-        --max-duration 600 \
+    --max-duration 600 \
-        --decoding-method fast_beam_search \
+    --decoding-method fast_beam_search \
-        --beam 4 \
+    --beam 4 \
-        --max-contexts 4 \
+    --max-contexts 4 \
-        --max-states 8
+    --max-states 8
 """
 
 
@@ -76,12 +76,17 @@ from beam_search import (
 )
 from gigaspeech_scoring import asr_text_post_processing
 from train import get_params, get_transducer_model
-
+from icefall.checkpoint import (
-from icefall.checkpoint import average_checkpoints, find_checkpoints, load_checkpoint
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
 from icefall.utils import (
     AttributeDict,
     setup_logger,
     store_transcripts,
+    str2bool,
     write_error_stats,
 )
 
@@ -94,9 +99,9 @@ def get_parser():
     parser.add_argument(
         "--epoch",
         type=int,
-        default=29,
+        default=30,
         help="""It specifies the checkpoint to use for decoding.
-        Note: Epoch counts from 0.
+        Note: Epoch counts from 1.
         You can specify --avg to use more checkpoints for model averaging.""",
     )
 
@@ -119,6 +124,17 @@ def get_parser():
         "'--epoch' and '--iter'",
     )
 
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=False,
+        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,
@@ -464,6 +480,9 @@ def main():
         params.suffix += f"-context-{params.context_size}"
         params.suffix += f"-max-sym-per-frame-{params.max_sym_per_frame}"
 
+    if params.use_averaged_model:
+        params.suffix += "-use-averaged-model"
+
     setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
     logging.info("Decoding started")
 
@@ -476,7 +495,7 @@ def main():
     sp = spm.SentencePieceProcessor()
     sp.load(params.bpe_model)
 
-    # <blk> and <unk> is defined in local/train_bpe_model.py
+    # <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()
@@ -486,37 +505,85 @@ def main():
     logging.info("About to create model")
     model = get_transducer_model(params)
 
-    if params.iter > 0:
+    if not params.use_averaged_model:
-        filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+        if params.iter > 0:
-            : params.avg
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
-        ]
+                : params.avg
-        if len(filenames) == 0:
+            ]
-            raise ValueError(
+            if len(filenames) == 0:
-                f"No checkpoints found for --iter {params.iter}, --avg {params.avg}"
+                raise ValueError(
-            )
+                    f"No checkpoints found for"
-        elif len(filenames) < params.avg:
+                    f" --iter {params.iter}, --avg {params.avg}"
-            raise ValueError(
+                )
-                f"Not enough checkpoints ({len(filenames)}) found for"
+            elif len(filenames) < params.avg:
-                f" --iter {params.iter}, --avg {params.avg}"
+                raise ValueError(
-            )
+                    f"Not enough checkpoints ({len(filenames)}) found for"
-        logging.info(f"averaging {filenames}")
+                    f" --iter {params.iter}, --avg {params.avg}"
-        model.to(device)
+                )
-        model.load_state_dict(average_checkpoints(filenames, device=device))
+            logging.info(f"averaging {filenames}")
-    elif params.avg == 1:
+            model.to(device)
-        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+            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:
-        start = params.epoch - params.avg + 1
+        if params.iter > 0:
-        filenames = []
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
-        for i in range(start, params.epoch + 1):
+                : params.avg + 1
-            if start >= 0:
+            ]
-                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            if len(filenames) == 0:
-        logging.info(f"averaging {filenames}")
+                raise ValueError(
-        model.to(device)
+                    f"No checkpoints found for"
-        model.load_state_dict(average_checkpoints(filenames, device=device))
+                    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
 
     if params.decoding_method == "fast_beam_search":
         decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)

egs/gigaspeech/ASR/pruned_transducer_stateless2/train.py

@@ -19,31 +19,30 @@
 """
 Usage:
 
-export CUDA_VISIBLE_DEVICES="0,1,2,3"
+export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"
 
 ./pruned_transducer_stateless2/train.py \
-  --world-size 4 \
+  --world-size 8 \
   --num-epochs 30 \
   --start-epoch 0 \
   --exp-dir pruned_transducer_stateless2/exp \
-  --full-libri 1 \
+  --max-duration 120
-  --max-duration 300
 
 # For mix precision training:
 
 ./pruned_transducer_stateless2/train.py \
-  --world-size 4 \
+  --world-size 8 \
   --num-epochs 30 \
   --start-epoch 0 \
   --use_fp16 1 \
   --exp-dir pruned_transducer_stateless2/exp \
-  --full-libri 1 \
+  --max-duration 200
-  --max-duration 550
 
 """
 
 
 import argparse
+import copy
 import logging
 import warnings
 from pathlib import Path
@@ -72,7 +71,10 @@ from torch.utils.tensorboard import SummaryWriter
 from icefall import diagnostics
 from icefall.checkpoint import load_checkpoint, remove_checkpoints
 from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
-from icefall.checkpoint import save_checkpoint_with_global_batch_idx
+from icefall.checkpoint import (
+    save_checkpoint_with_global_batch_idx,
+    update_averaged_model,
+)
 from icefall.dist import cleanup_dist, setup_dist
 from icefall.env import get_env_info
 from icefall.utils import AttributeDict, MetricsTracker, setup_logger, str2bool
@@ -116,10 +118,10 @@ def get_parser():
     parser.add_argument(
         "--start-epoch",
         type=int,
-        default=0,
+        default=1,
-        help="""Resume training from from this epoch.
+        help="""Resume training from this epoch.
-        If it is positive, it will load checkpoint from
+        If larger than 1, it will load checkpoint from
-        transducer_stateless2/exp/epoch-{start_epoch-1}.pt
+        exp-dir/epoch-{start_epoch-1}.pt
         """,
     )
 
@@ -242,7 +244,7 @@ def get_parser():
     parser.add_argument(
         "--keep-last-k",
         type=int,
-        default=20,
+        default=30,
         help="""Only keep this number of checkpoints on disk.
         For instance, if it is 3, there are only 3 checkpoints
         in the exp-dir with filenames `checkpoint-xxx.pt`.
@@ -250,6 +252,19 @@ def get_parser():
         """,
     )
 
+    parser.add_argument(
+        "--average-period",
+        type=int,
+        default=200,
+        help="""Update the averaged model, namely `model_avg`, after processing
+        this number of batches. `model_avg` is a separate version of model,
+        in which each floating-point parameter is the average of all the
+        parameters from the start of training. Each time we take the average,
+        we do: `model_avg = model * (average_period / batch_idx_train) +
+            model_avg * ((batch_idx_train - average_period) / batch_idx_train)`.
+        """,
+    )
+
     parser.add_argument(
         "--use-fp16",
         type=str2bool,
@@ -387,6 +402,7 @@ def get_transducer_model(params: AttributeDict) -> nn.Module:
 def load_checkpoint_if_available(
     params: AttributeDict,
     model: nn.Module,
+    model_avg: nn.Module = None,
     optimizer: Optional[torch.optim.Optimizer] = None,
     scheduler: Optional[LRSchedulerType] = None,
 ) -> Optional[Dict[str, Any]]:
@@ -394,7 +410,7 @@ def load_checkpoint_if_available(
 
     If params.start_batch is positive, it will load the checkpoint from
     `params.exp_dir/checkpoint-{params.start_batch}.pt`. Otherwise, if
-    params.start_epoch is positive, it will load the checkpoint from
+    params.start_epoch is larger than 1, it will load the checkpoint from
     `params.start_epoch - 1`.
 
     Apart from loading state dict for `model` and `optimizer` it also updates
@@ -406,6 +422,8 @@ def load_checkpoint_if_available(
         The return value of :func:`get_params`.
       model:
         The training model.
+      model_avg:
+        The stored model averaged from the start of training.
       optimizer:
         The optimizer that we are using.
       scheduler:
@@ -415,7 +433,7 @@ def load_checkpoint_if_available(
     """
     if params.start_batch > 0:
         filename = params.exp_dir / f"checkpoint-{params.start_batch}.pt"
-    elif params.start_epoch > 0:
+    elif params.start_epoch > 1:
         filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
     else:
         return None
@@ -425,6 +443,7 @@ def load_checkpoint_if_available(
     saved_params = load_checkpoint(
         filename,
         model=model,
+        model_avg=model_avg,
         optimizer=optimizer,
         scheduler=scheduler,
     )
@@ -451,7 +470,8 @@ def load_checkpoint_if_available(
 
 def save_checkpoint(
     params: AttributeDict,
-    model: nn.Module,
+    model: Union[nn.Module, DDP],
+    model_avg: Optional[nn.Module] = None,
     optimizer: Optional[torch.optim.Optimizer] = None,
     scheduler: Optional[LRSchedulerType] = None,
     sampler: Optional[CutSampler] = None,
@@ -465,6 +485,8 @@ def save_checkpoint(
         It is returned by :func:`get_params`.
       model:
         The training model.
+      model_avg:
+        The stored model averaged from the start of training.
       optimizer:
         The optimizer used in the training.
       sampler:
@@ -478,6 +500,7 @@ def save_checkpoint(
     save_checkpoint_impl(
         filename=filename,
         model=model,
+        model_avg=model_avg,
         params=params,
         optimizer=optimizer,
         scheduler=scheduler,
@@ -497,14 +520,14 @@ def save_checkpoint(
 
 def compute_loss(
     params: AttributeDict,
-    model: nn.Module,
+    model: Union[nn.Module, DDP],
     sp: spm.SentencePieceProcessor,
     batch: dict,
     is_training: bool,
     warmup: float = 1.0,
 ) -> Tuple[Tensor, MetricsTracker]:
     """
-    Compute CTC loss given the model and its inputs.
+    Compute transducer loss given the model and its inputs.
 
     Args:
       params:
@@ -570,7 +593,7 @@ def compute_loss(
 
 def compute_validation_loss(
     params: AttributeDict,
-    model: nn.Module,
+    model: Union[nn.Module, DDP],
     sp: spm.SentencePieceProcessor,
     valid_dl: torch.utils.data.DataLoader,
     world_size: int = 1,
@@ -604,13 +627,14 @@ def compute_validation_loss(
 
 def train_one_epoch(
     params: AttributeDict,
-    model: nn.Module,
+    model: Union[nn.Module, DDP],
     optimizer: torch.optim.Optimizer,
     scheduler: LRSchedulerType,
     sp: spm.SentencePieceProcessor,
     train_dl: torch.utils.data.DataLoader,
     valid_dl: torch.utils.data.DataLoader,
     scaler: GradScaler,
+    model_avg: Optional[nn.Module] = None,
     tb_writer: Optional[SummaryWriter] = None,
     world_size: int = 1,
     rank: int = 0,
@@ -636,6 +660,8 @@ def train_one_epoch(
         Dataloader for the validation dataset.
       scaler:
         The scaler used for mix precision training.
+      model_avg:
+        The stored model averaged from the start of training.
       tb_writer:
         Writer to write log messages to tensorboard.
       world_size:
@@ -662,6 +688,7 @@ def train_one_epoch(
             loss, loss_info = compute_loss(
                 params=params,
                 model=model,
+                model_avg=model_avg,
                 sp=sp,
                 batch=batch,
                 is_training=True,
@@ -690,6 +717,7 @@ def train_one_epoch(
                 out_dir=params.exp_dir,
                 global_batch_idx=params.batch_idx_train,
                 model=model,
+                model_avg=model_avg,
                 params=params,
                 optimizer=optimizer,
                 scheduler=scheduler,
@@ -793,7 +821,16 @@ def run(rank, world_size, args):
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    checkpoints = load_checkpoint_if_available(params=params, model=model)
+    assert params.save_every_n >= params.average_period
+    model_avg: Optional[nn.Module] = None
+    if rank == 0:
+        # model_avg is only used with rank 0
+        model_avg = copy.deepcopy(model).to(torch.float64)
+
+    assert params.start_epoch > 0, params.start_epoch
+    checkpoints = load_checkpoint_if_available(
+        params=params, model=model, model_avg=model_avg
+    )
 
     model.to(device)
     if world_size > 1:
@@ -852,10 +889,10 @@ def run(rank, world_size, args):
         logging.info("Loading grad scaler state dict")
         scaler.load_state_dict(checkpoints["grad_scaler"])
 
-    for epoch in range(params.start_epoch, params.num_epochs):
+    for epoch in range(params.start_epoch, params.num_epochs + 1):
-        scheduler.step_epoch(epoch)
+        scheduler.step_epoch(epoch - 1)
-        fix_random_seed(params.seed + epoch)
+        fix_random_seed(params.seed + epoch - 1)
-        train_dl.sampler.set_epoch(epoch)
+        train_dl.sampler.set_epoch(epoch - 1)
 
         if tb_writer is not None:
             tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
@@ -865,6 +902,7 @@ def run(rank, world_size, args):
         train_one_epoch(
             params=params,
             model=model,
+            model_avg=model_avg,
             optimizer=optimizer,
             scheduler=scheduler,
             sp=sp,
@@ -883,6 +921,7 @@ def run(rank, world_size, args):
         save_checkpoint(
             params=params,
             model=model,
+            model_avg=model_avg,
             optimizer=optimizer,
             scheduler=scheduler,
             sampler=train_dl.sampler,
@@ -898,7 +937,7 @@ def run(rank, world_size, args):
 
 
 def scan_pessimistic_batches_for_oom(
-    model: nn.Module,
+    model: Union[nn.Module, DDP],
     train_dl: torch.utils.data.DataLoader,
     optimizer: torch.optim.Optimizer,
     sp: spm.SentencePieceProcessor,

egs/librispeech/ASR/local/compute_fbank_librispeech.py

@@ -121,10 +121,10 @@ def compute_fbank_librispeech(
                 recordings=m["recordings"],
                 supervisions=m["supervisions"],
             )
-            if bpe_model:
-                cut_set = filter_cuts(cut_set, sp)
 
             if "train" in partition:
+                if bpe_model:
+                    cut_set = filter_cuts(cut_set, sp)
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )

egs/librispeech/ASR/lstm_transducer_stateless2/decode.py

@@ -550,7 +550,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LM=LM,
         )
         for hyp in sp.decode(hyp_tokens):
@@ -561,7 +560,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LODR_lm=ngram_lm,
             LODR_lm_scale=ngram_lm_scale,
             LM=LM,

egs/librispeech/ASR/lstm_transducer_stateless2/export-onnx-zh.py

@@ -0,0 +1,632 @@
+#!/usr/bin/env python3
+#
+# Copyright 2023 Xiaomi Corporation (Author: Fangjun Kuang)
+
+"""
+This script exports a transducer model from PyTorch to ONNX.
+
+We use the pre-trained model from
+https://huggingface.co/csukuangfj/icefall-asr-wenetspeech-lstm-transducer-stateless-2022-10-14
+as an example to show how to use this file.
+
+1. Download the pre-trained model
+
+cd egs/librispeech/ASR
+
+repo_url=https://huggingface.co/csukuangfj/icefall-asr-wenetspeech-lstm-transducer-stateless-2022-10-14
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+repo=$(basename $repo_url)
+
+pushd $repo
+git lfs pull --include "data/lexicon.txt"
+git lfs pull --include "data/L.pt"
+git lfs pull --include "exp/epoch-11.pt"
+git lfs pull --include "exp/epoch-10.pt"
+
+popd
+
+2. Export the model to ONNX
+
+./lstm_transducer_stateless2/export-onnx-zh.py \
+  --lang-dir ./icefall-asr-wenetspeech-lstm-transducer-stateless-2022-10-14/data/lang_char \
+  --use-averaged-model 1 \
+  --epoch 11 \
+  --avg 1 \
+  --exp-dir ./icefall-asr-wenetspeech-lstm-transducer-stateless-2022-10-14/exp \
+  --num-encoder-layers 12 \
+  --encoder-dim 512 \
+  --rnn-hidden-size 1024
+
+It will generate the following files inside $repo/exp:
+
+  - encoder-epoch-11-avg-1.onnx
+  - decoder-epoch-11-avg-1.onnx
+  - joiner-epoch-11-avg-1.onnx
+  - encoder-epoch-11-avg-1.int8.onnx
+  - decoder-epoch-11-avg-1.int8.onnx
+  - joiner-epoch-11-avg-1.int8.onnx
+
+See ./onnx_pretrained.py and ./onnx_check.py for how to
+use the exported ONNX models.
+"""
+
+import argparse
+import logging
+from pathlib import Path
+from typing import Dict, Optional, Tuple
+
+import onnx
+import torch
+import torch.nn as nn
+from decoder import Decoder
+from lstm import RNN
+from onnxruntime.quantization import QuantType, quantize_dynamic
+from scaling_converter import convert_scaled_to_non_scaled
+from train import add_model_arguments, get_params, get_transducer_model
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.lexicon import Lexicon
+from icefall.utils import setup_logger, str2bool
+
+
+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 averaging.
+        Note: Epoch counts from 0.
+        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_stateless5/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_char",
+        help="The lang dir",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def add_meta_data(filename: str, meta_data: Dict[str, str]):
+    """Add meta data to an ONNX model. It is changed in-place.
+
+    Args:
+      filename:
+        Filename of the ONNX model to be changed.
+      meta_data:
+        Key-value pairs.
+    """
+    model = onnx.load(filename)
+    for key, value in meta_data.items():
+        meta = model.metadata_props.add()
+        meta.key = key
+        meta.value = value
+
+    onnx.save(model, filename)
+
+
+class OnnxEncoder(nn.Module):
+    """A wrapper for RNN and the encoder_proj from the joiner"""
+
+    def __init__(self, encoder: RNN, encoder_proj: nn.Linear):
+        """
+        Args:
+          encoder:
+            An RNN encoder.
+          encoder_proj:
+            The projection layer for encoder from the joiner.
+        """
+        super().__init__()
+        self.encoder = encoder
+        self.encoder_proj = encoder_proj
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        states: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """Please see the help information of RNN.forward
+
+        Args:
+          x:
+            A 3-D tensor of shape (N, T, C)
+          states:
+            A tuple of 2 tensors (optional). It is for streaming inference.
+            states[0] is the hidden states of all layers,
+              with shape of (num_layers, N, d_model);
+            states[1] is the cell states of all layers,
+              with shape of (num_layers, N, rnn_hidden_size).
+        Returns:
+          Return a tuple containing:
+            - encoder_out, A 3-D tensor of shape (N, T', joiner_dim)
+            - updated states, whose shape is the same as the input states.
+        """
+        N = x.size(0)
+        T = x.size(1)
+        x_lens = torch.tensor([T] * N, dtype=torch.int64, device=x.device)
+        encoder_out, _, next_states = self.encoder(x, x_lens, states)
+
+        encoder_out = self.encoder_proj(encoder_out)
+        # Now encoder_out is of shape (N, T, joiner_dim)
+
+        return encoder_out, next_states
+
+
+class OnnxDecoder(nn.Module):
+    """A wrapper for Decoder and the decoder_proj from the joiner"""
+
+    def __init__(self, decoder: Decoder, decoder_proj: nn.Linear):
+        super().__init__()
+        self.decoder = decoder
+        self.decoder_proj = decoder_proj
+
+    def forward(self, y: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, context_size).
+        Returns
+          Return a 2-D tensor of shape (N, joiner_dim)
+        """
+        need_pad = False
+        decoder_output = self.decoder(y, need_pad=need_pad)
+        decoder_output = decoder_output.squeeze(1)
+        output = self.decoder_proj(decoder_output)
+
+        return output
+
+
+class OnnxJoiner(nn.Module):
+    """A wrapper for the joiner"""
+
+    def __init__(self, output_linear: nn.Linear):
+        super().__init__()
+        self.output_linear = output_linear
+
+    def forward(
+        self,
+        encoder_out: torch.Tensor,
+        decoder_out: torch.Tensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+          decoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+        Returns:
+          Return a 2-D tensor of shape (N, vocab_size)
+        """
+        logit = encoder_out + decoder_out
+        logit = self.output_linear(torch.tanh(logit))
+        return logit
+
+
+def export_encoder_model_onnx(
+    encoder_model: OnnxEncoder,
+    encoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the given encoder model to ONNX format.
+    The exported model has the following inputs:
+
+        - x, a tensor of shape (N, T, C); dtype is torch.float32
+        - state0, a tensor of shape (num_encoder_layers, batch_size, d_model)
+        - state1, a tensor of shape (num_encoder_layers, batch_size, rnn_hidden_size)
+
+    and it has 3 outputs:
+
+        - encoder_out, a tensor of shape (N, T', joiner_dim)
+        - new_state0, a tensor of shape (num_encoder_layers, batch_size, d_model)
+        - new_state1, a tensor of shape (num_encoder_layers, batch_size, rnn_hidden_size)
+
+    Args:
+      encoder_model:
+        The input encoder model
+      encoder_filename:
+        The filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+    num_encoder_layers = encoder_model.encoder.num_encoder_layers
+    d_model = encoder_model.encoder.d_model
+    rnn_hidden_size = encoder_model.encoder.rnn_hidden_size
+
+    decode_chunk_len = 4
+    T = 9
+
+    x = torch.zeros(1, T, 80, dtype=torch.float32)
+    states = encoder_model.encoder.get_init_states()
+    # state0: (num_encoder_layers, batch_size, d_model)
+    # state1: (num_encoder_layers, batch_size, rnn_hidden_size)
+
+    torch.onnx.export(
+        encoder_model,
+        (x, states),
+        encoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=["x", "state0", "state1"],
+        output_names=["encoder_out", "new_state0", "new_state1"],
+        dynamic_axes={
+            "x": {0: "N", 1: "T"},
+            "state0": {1: "N"},
+            "state1": {1: "N"},
+            "encoder_out": {0: "N"},
+            "new_state0": {1: "N"},
+            "new_state1": {1: "N"},
+        },
+    )
+
+    meta_data = {
+        "model_type": "lstm",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "decode_chunk_len": str(decode_chunk_len),  # 32
+        "T": str(T),  # 39
+        "num_encoder_layers": str(num_encoder_layers),
+        "d_model": str(d_model),
+        "rnn_hidden_size": str(rnn_hidden_size),
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
+
+def export_decoder_model_onnx(
+    decoder_model: OnnxDecoder,
+    decoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the decoder model to ONNX format.
+
+    The exported model has one input:
+
+        - y: a torch.int64 tensor of shape (N, decoder_model.context_size)
+
+    and has one output:
+
+        - decoder_out: a torch.float32 tensor of shape (N, joiner_dim)
+
+    Args:
+      decoder_model:
+        The decoder model to be exported.
+      decoder_filename:
+        Filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+    context_size = decoder_model.decoder.context_size
+    vocab_size = decoder_model.decoder.vocab_size
+
+    y = torch.zeros(10, context_size, dtype=torch.int64)
+    torch.onnx.export(
+        decoder_model,
+        y,
+        decoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=["y"],
+        output_names=["decoder_out"],
+        dynamic_axes={
+            "y": {0: "N"},
+            "decoder_out": {0: "N"},
+        },
+    )
+
+    meta_data = {
+        "context_size": str(context_size),
+        "vocab_size": str(vocab_size),
+    }
+    add_meta_data(filename=decoder_filename, meta_data=meta_data)
+
+
+def export_joiner_model_onnx(
+    joiner_model: nn.Module,
+    joiner_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the joiner model to ONNX format.
+    The exported joiner model has two inputs:
+
+        - encoder_out: a tensor of shape (N, joiner_dim)
+        - decoder_out: a tensor of shape (N, joiner_dim)
+
+    and produces one output:
+
+        - logit: a tensor of shape (N, vocab_size)
+    """
+    joiner_dim = joiner_model.output_linear.weight.shape[1]
+    logging.info(f"joiner dim: {joiner_dim}")
+
+    projected_encoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+    projected_decoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+
+    torch.onnx.export(
+        joiner_model,
+        (projected_encoder_out, projected_decoder_out),
+        joiner_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=[
+            "encoder_out",
+            "decoder_out",
+        ],
+        output_names=["logit"],
+        dynamic_axes={
+            "encoder_out": {0: "N"},
+            "decoder_out": {0: "N"},
+            "logit": {0: "N"},
+        },
+    )
+    meta_data = {
+        "joiner_dim": str(joiner_dim),
+    }
+    add_meta_data(filename=joiner_filename, meta_data=meta_data)
+
+
+@torch.no_grad()
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    setup_logger(f"{params.exp_dir}/log-export/log-export-onnx")
+
+    logging.info(f"device: {device}")
+
+    lexicon = Lexicon(params.lang_dir)
+    params.blank_id = 0
+    params.vocab_size = max(lexicon.tokens) + 1
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params, enable_giga=False)
+
+    model.to(device)
+
+    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), strict=False
+            )
+        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), strict=False
+            )
+    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,
+                ),
+                strict=False,
+            )
+        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,
+                ),
+                strict=False,
+            )
+
+    model.to("cpu")
+    model.eval()
+
+    convert_scaled_to_non_scaled(model, inplace=True, is_onnx=True)
+
+    encoder = OnnxEncoder(
+        encoder=model.encoder,
+        encoder_proj=model.joiner.encoder_proj,
+    )
+
+    decoder = OnnxDecoder(
+        decoder=model.decoder,
+        decoder_proj=model.joiner.decoder_proj,
+    )
+
+    joiner = OnnxJoiner(output_linear=model.joiner.output_linear)
+
+    encoder_num_param = sum([p.numel() for p in encoder.parameters()])
+    decoder_num_param = sum([p.numel() for p in decoder.parameters()])
+    joiner_num_param = sum([p.numel() for p in joiner.parameters()])
+    total_num_param = encoder_num_param + decoder_num_param + joiner_num_param
+    logging.info(f"encoder parameters: {encoder_num_param}")
+    logging.info(f"decoder parameters: {decoder_num_param}")
+    logging.info(f"joiner parameters: {joiner_num_param}")
+    logging.info(f"total parameters: {total_num_param}")
+
+    if params.iter > 0:
+        suffix = f"iter-{params.iter}"
+    else:
+        suffix = f"epoch-{params.epoch}"
+
+    suffix += f"-avg-{params.avg}"
+
+    opset_version = 13
+
+    logging.info("Exporting encoder")
+    encoder_filename = params.exp_dir / f"encoder-{suffix}.onnx"
+    export_encoder_model_onnx(
+        encoder,
+        encoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported encoder to {encoder_filename}")
+
+    logging.info("Exporting decoder")
+    decoder_filename = params.exp_dir / f"decoder-{suffix}.onnx"
+    export_decoder_model_onnx(
+        decoder,
+        decoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported decoder to {decoder_filename}")
+
+    logging.info("Exporting joiner")
+    joiner_filename = params.exp_dir / f"joiner-{suffix}.onnx"
+    export_joiner_model_onnx(
+        joiner,
+        joiner_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported joiner to {joiner_filename}")
+
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    main()

egs/librispeech/ASR/lstm_transducer_stateless2/export-onnx.py

@@ -34,11 +34,14 @@ popd
   --avg 1 \
   --exp-dir $repo/exp
 
-It will generate the following 3 files inside $repo/exp:
+It will generate the following files inside $repo/exp:
 
   - encoder-epoch-99-avg-1.onnx
   - decoder-epoch-99-avg-1.onnx
   - joiner-epoch-99-avg-1.onnx
+  - encoder-epoch-99-avg-1.int8.onnx
+  - decoder-epoch-99-avg-1.int8.onnx
+  - joiner-epoch-99-avg-1.int8.onnx
 
 See ./onnx_pretrained.py and ./onnx_check.py for how to
 use the exported ONNX models.
@@ -55,6 +58,7 @@ import torch
 import torch.nn as nn
 from decoder import Decoder
 from lstm import RNN
+from onnxruntime.quantization import QuantType, quantize_dynamic
 from scaling_converter import convert_scaled_to_non_scaled
 from train import add_model_arguments, get_params, get_transducer_model
 
@@ -586,6 +590,35 @@ def main():
     )
     logging.info(f"Exported joiner to {joiner_filename}")
 
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
 
 if __name__ == "__main__":
     formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"

egs/librispeech/ASR/pruned_transducer_stateless/decode.py

@@ -107,6 +107,7 @@ Usage:
 
 import argparse
 import logging
+import math
 from collections import defaultdict
 from pathlib import Path
 from typing import Dict, List, Optional, Tuple
@@ -138,6 +139,8 @@ from icefall.utils import (
     write_error_stats,
 )
 
+LOG_EPS = math.log(1e-10)
+
 
 def get_parser():
     parser = argparse.ArgumentParser(
@@ -288,7 +291,7 @@ def get_parser():
         "--decode-chunk-size",
         type=int,
         default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
+        help="The chunk size for decoding (in frames after subsampling). Set -1 to use full attention.",
     )
     parser.add_argument(
         "--left-context",
@@ -370,6 +373,14 @@ def decode_one_batch(
     feature_lens = supervisions["num_frames"].to(device)
 
     if params.simulate_streaming:
+        if params.decode_chunk_size > 0:
+            # except the case of using full attention
+            feature_lens += params.left_context
+            feature = torch.nn.functional.pad(
+                feature,
+                pad=(0, 0, 0, params.left_context),
+                value=LOG_EPS,
+            )
         encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
             x=feature,
             x_lens=feature_lens,

egs/librispeech/ASR/pruned_transducer_stateless2/beam_search.py

@@ -829,11 +829,22 @@ class HypothesisList(object):
                 ans.add(hyp)  # shallow copy
         return ans
 
-    def topk(self, k: int) -> "HypothesisList":
+    def topk(self, k: int, length_norm: bool = False) -> "HypothesisList":
-        """Return the top-k hypothesis."""
+        """Return the top-k hypothesis.
+
+        Args:
+          length_norm:
+            If True, the `log_prob` of a hypothesis is normalized by the
+            number of tokens in it.
+        """
         hyps = list(self._data.items())
 
-        hyps = sorted(hyps, key=lambda h: h[1].log_prob, reverse=True)[:k]
+        if length_norm:
+            hyps = sorted(
+                hyps, key=lambda h: h[1].log_prob / len(h[1].ys), reverse=True
+            )[:k]
+        else:
+            hyps = sorted(hyps, key=lambda h: h[1].log_prob, reverse=True)[:k]
 
         ans = HypothesisList(dict(hyps))
         return ans
@@ -1852,7 +1863,6 @@ def modified_beam_search_LODR(
     model: Transducer,
     encoder_out: torch.Tensor,
     encoder_out_lens: torch.Tensor,
-    sp: spm.SentencePieceProcessor,
     LODR_lm: NgramLm,
     LODR_lm_scale: float,
     LM: LmScorer,
@@ -1872,8 +1882,6 @@ def modified_beam_search_LODR(
         encoder_out_lens (torch.Tensor):
             A 1-D tensor of shape (N,), containing the number of
             valid frames in encoder_out before padding.
-        sp:
-            Sentence piece generator.
         LODR_lm:
             A low order n-gram LM, whose score will be subtracted during shallow fusion
         LODR_lm_scale:
@@ -1901,7 +1909,7 @@ def modified_beam_search_LODR(
     )
 
     blank_id = model.decoder.blank_id
-    sos_id = sp.piece_to_id("<sos/eos>")
+    sos_id = getattr(LM, "sos_id", 1)
     unk_id = getattr(model, "unk_id", blank_id)
     context_size = model.decoder.context_size
     device = next(model.parameters()).device
@@ -2126,7 +2134,6 @@ def modified_beam_search_lm_shallow_fusion(
     model: Transducer,
     encoder_out: torch.Tensor,
     encoder_out_lens: torch.Tensor,
-    sp: spm.SentencePieceProcessor,
     LM: LmScorer,
     beam: int = 4,
     return_timestamps: bool = False,
@@ -2165,7 +2172,7 @@ def modified_beam_search_lm_shallow_fusion(
     )
 
     blank_id = model.decoder.blank_id
-    sos_id = sp.piece_to_id("<sos/eos>")
+    sos_id = getattr(LM, "sos_id", 1)
     unk_id = getattr(model, "unk_id", blank_id)
     context_size = model.decoder.context_size
     device = next(model.parameters()).device

egs/librispeech/ASR/pruned_transducer_stateless2/conformer.py

@@ -375,6 +375,11 @@ class Conformer(EncoderInterface):
 
             assert x.size(0) == lengths.max().item()
 
+            if chunk_size < 0:
+                # use full attention
+                chunk_size = x.size(0)
+                left_context = -1
+
             num_left_chunks = -1
             if left_context >= 0:
                 assert left_context % chunk_size == 0

egs/librispeech/ASR/pruned_transducer_stateless2/decode.py

@@ -295,7 +295,7 @@ def get_parser():
         "--decode-chunk-size",
         type=int,
         default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
+        help="The chunk size for decoding (in frames after subsampling). Set -1 to use full attention.",
     )
 
     parser.add_argument(
@@ -378,12 +378,14 @@ def decode_one_batch(
     feature_lens = supervisions["num_frames"].to(device)
 
     if params.simulate_streaming:
-        feature_lens += params.left_context
+        if params.decode_chunk_size > 0:
-        feature = torch.nn.functional.pad(
+            # except the case of using full attention
-            feature,
+            feature_lens += params.left_context
-            pad=(0, 0, 0, params.left_context),
+            feature = torch.nn.functional.pad(
-            value=LOG_EPS,
+                feature,
-        )
+                pad=(0, 0, 0, params.left_context),
+                value=LOG_EPS,
+            )
         encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
             x=feature,
             x_lens=feature_lens,

egs/librispeech/ASR/pruned_transducer_stateless3/decode.py

@@ -344,7 +344,7 @@ def get_parser():
         "--decode-chunk-size",
         type=int,
         default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
+        help="The chunk size for decoding (in frames after subsampling). Set -1 to use full attention.",
     )
 
     parser.add_argument(
@@ -508,12 +508,14 @@ def decode_one_batch(
     feature_lens = supervisions["num_frames"].to(device)
 
     if params.simulate_streaming:
-        feature_lens += params.left_context
+        if params.decode_chunk_size > 0:
-        feature = torch.nn.functional.pad(
+            # except the case of using full attention
-            feature,
+            feature_lens += params.left_context
-            pad=(0, 0, 0, params.left_context),
+            feature = torch.nn.functional.pad(
-            value=LOG_EPS,
+                feature,
-        )
+                pad=(0, 0, 0, params.left_context),
+                value=LOG_EPS,
+            )
         encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
             x=feature,
             x_lens=feature_lens,
@@ -673,7 +675,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LM=LM,
         )
         for hyp in sp.decode(hyp_tokens):
@@ -684,7 +685,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LODR_lm=ngram_lm,
             LODR_lm_scale=ngram_lm_scale,
             LM=LM,

egs/librispeech/ASR/pruned_transducer_stateless3/export-onnx.py

@@ -54,6 +54,7 @@ import torch
 import torch.nn as nn
 from conformer import Conformer
 from decoder import Decoder
+from onnxruntime.quantization import QuantType, quantize_dynamic
 from scaling_converter import convert_scaled_to_non_scaled
 from train import add_model_arguments, get_params, get_transducer_model
 
@@ -273,6 +274,16 @@ def export_encoder_model_onnx(
         },
     )
 
+    meta_data = {
+        "model_type": "conformer",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "comment": "stateless3",
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
 
 def export_decoder_model_onnx(
     decoder_model: OnnxDecoder,
@@ -490,6 +501,35 @@ def main():
     )
     logging.info(f"Exported joiner to {joiner_filename}")
 
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
 
 if __name__ == "__main__":
     formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"

egs/librispeech/ASR/pruned_transducer_stateless3/onnx_pretrained.py

@@ -403,9 +403,8 @@ def main():
             text += symbol_table[i]
         return text.replace("▁", " ").strip()
 
-    context_size = model.context_size
     for filename, hyp in zip(args.sound_files, hyps):
-        words = token_ids_to_words(hyp[context_size:])
+        words = token_ids_to_words(hyp)
         s += f"{filename}:\n{words}\n"
     logging.info(s)
 

egs/librispeech/ASR/pruned_transducer_stateless4/decode.py

@@ -326,14 +326,14 @@ def get_parser():
         "--decode-chunk-size",
         type=int,
         default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
+        help="The chunk size for decoding (in frames after subsampling). Set -1 to use full attention.",
     )
 
     parser.add_argument(
         "--left-context",
         type=int,
         default=64,
-        help="left context can be seen during decoding (in frames after subsampling)",  # noqa
+        help="""Left context can be seen during decoding (in frames after subsampling). """,  # noqa
     )
 
     parser.add_argument(
@@ -409,12 +409,14 @@ def decode_one_batch(
     feature_lens = supervisions["num_frames"].to(device)
 
     if params.simulate_streaming:
-        feature_lens += params.left_context
+        if params.decode_chunk_size > 0:
-        feature = torch.nn.functional.pad(
+            # except the case of using full attention
-            feature,
+            feature_lens += params.left_context
-            pad=(0, 0, 0, params.left_context),
+            feature = torch.nn.functional.pad(
-            value=LOG_EPS,
+                feature,
-        )
+                pad=(0, 0, 0, params.left_context),
+                value=LOG_EPS,
+            )
         encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
             x=feature,
             x_lens=feature_lens,

egs/librispeech/ASR/pruned_transducer_stateless5/decode.py

@@ -291,7 +291,7 @@ def get_parser():
         "--decode-chunk-size",
         type=int,
         default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
+        help="The chunk size for decoding (in frames after subsampling). Set -1 to use full attention.",
     )
 
     parser.add_argument(
@@ -470,12 +470,14 @@ def decode_one_batch(
     feature_lens = supervisions["num_frames"].to(device)
 
     if params.simulate_streaming:
-        feature_lens += params.left_context
+        if params.decode_chunk_size > 0:
-        feature = torch.nn.functional.pad(
+            # except the case of using full attention
-            feature,
+            feature_lens += params.left_context
-            pad=(0, 0, 0, params.left_context),
+            feature = torch.nn.functional.pad(
-            value=LOG_EPS,
+                feature,
-        )
+                pad=(0, 0, 0, params.left_context),
+                value=LOG_EPS,
+            )
         encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
             x=feature,
             x_lens=feature_lens,
@@ -584,7 +586,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LM=LM,
         )
         for hyp in sp.decode(hyp_tokens):
@@ -595,7 +596,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LODR_lm=ngram_lm,
             LODR_lm_scale=ngram_lm_scale,
             LM=LM,

egs/librispeech/ASR/pruned_transducer_stateless5/export-onnx.py

@@ -296,6 +296,16 @@ def export_encoder_model_onnx(
         },
     )
 
+    meta_data = {
+        "model_type": "conformer",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "comment": "stateless5",
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
 
 def export_decoder_model_onnx(
     decoder_model: OnnxDecoder,

egs/librispeech/ASR/pruned_transducer_stateless7/alignment.py

@@ -0,0 +1,206 @@
+# Copyright    2022-2023  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                       Zengwei Yao)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+from typing import List
+
+import k2
+import torch
+
+from beam_search import Hypothesis, HypothesisList, get_hyps_shape
+
+# The force alignment problem can be formulated as finding
+# a path in a rectangular lattice, where the path starts
+# from the lower left corner and ends at the upper right
+# corner. The horizontal axis of the lattice is `t` (representing
+# acoustic frame indexes) and the vertical axis is `u` (representing
+# BPE tokens of the transcript).
+#
+# The notations `t` and `u` are from the paper
+# https://arxiv.org/pdf/1211.3711.pdf
+#
+# Beam search is used to find the path with the highest log probabilities.
+#
+# It assumes the maximum number of symbols that can be
+# emitted per frame is 1.
+
+
+def batch_force_alignment(
+    model: torch.nn.Module,
+    encoder_out: torch.Tensor,
+    encoder_out_lens: torch.Tensor,
+    ys_list: List[List[int]],
+    beam_size: int = 4,
+) -> List[int]:
+    """Compute the force alignment of a batch of utterances given their transcripts
+    in BPE tokens and the corresponding acoustic output from the encoder.
+
+    Caution:
+      This function is modified from `modified_beam_search` in beam_search.py.
+      We assume that the maximum number of sybmols per frame is 1.
+
+    Args:
+      model:
+        The transducer model.
+      encoder_out:
+        A tensor of shape (N, T, C).
+      encoder_out_lens:
+        A 1-D tensor of shape (N,), containing number of valid frames in
+        encoder_out before padding.
+      ys_list:
+        A list of BPE token IDs list. We require that for each utterance i,
+        len(ys_list[i]) <= encoder_out_lens[i].
+      beam_size:
+        Size of the beam used in beam search.
+
+    Returns:
+      Return a list of frame indexes list for each utterance i,
+      where len(ans[i]) == len(ys_list[i]).
+    """
+    assert encoder_out.ndim == 3, encoder_out.ndim
+    assert encoder_out.size(0) == len(ys_list), (encoder_out.size(0), len(ys_list))
+    assert encoder_out.size(0) > 0, encoder_out.size(0)
+
+    blank_id = model.decoder.blank_id
+    context_size = model.decoder.context_size
+    device = next(model.parameters()).device
+
+    packed_encoder_out = torch.nn.utils.rnn.pack_padded_sequence(
+        input=encoder_out,
+        lengths=encoder_out_lens.cpu(),
+        batch_first=True,
+        enforce_sorted=False,
+    )
+    batch_size_list = packed_encoder_out.batch_sizes.tolist()
+    N = encoder_out.size(0)
+    assert torch.all(encoder_out_lens > 0), encoder_out_lens
+    assert N == batch_size_list[0], (N, batch_size_list)
+
+    sorted_indices = packed_encoder_out.sorted_indices.tolist()
+    encoder_out_lens = encoder_out_lens.tolist()
+    ys_lens = [len(ys) for ys in ys_list]
+    sorted_encoder_out_lens = [encoder_out_lens[i] for i in sorted_indices]
+    sorted_ys_lens = [ys_lens[i] for i in sorted_indices]
+    sorted_ys_list = [ys_list[i] for i in sorted_indices]
+
+    B = [HypothesisList() for _ in range(N)]
+    for i in range(N):
+        B[i].add(
+            Hypothesis(
+                ys=[blank_id] * context_size,
+                log_prob=torch.zeros(1, dtype=torch.float32, device=device),
+                timestamp=[],
+            )
+        )
+
+    encoder_out = model.joiner.encoder_proj(packed_encoder_out.data)
+
+    offset = 0
+    finalized_B = []
+    for (t, batch_size) in enumerate(batch_size_list):
+        start = offset
+        end = offset + batch_size
+        current_encoder_out = encoder_out.data[start:end]
+        current_encoder_out = current_encoder_out.unsqueeze(1).unsqueeze(1)
+        # current_encoder_out's shape is (batch_size, 1, 1, encoder_out_dim)
+        offset = end
+
+        finalized_B = B[batch_size:] + finalized_B
+        B = B[:batch_size]
+        sorted_encoder_out_lens = sorted_encoder_out_lens[:batch_size]
+        sorted_ys_lens = sorted_ys_lens[:batch_size]
+
+        hyps_shape = get_hyps_shape(B).to(device)
+
+        A = [list(b) for b in B]
+        B = [HypothesisList() for _ in range(batch_size)]
+
+        ys_log_probs = torch.cat(
+            [hyp.log_prob.reshape(1, 1) for hyps in A for hyp in hyps]
+        )  # (num_hyps, 1)
+
+        decoder_input = torch.tensor(
+            [hyp.ys[-context_size:] for hyps in A for hyp in hyps],
+            device=device,
+            dtype=torch.int64,
+        )  # (num_hyps, context_size)
+
+        decoder_out = model.decoder(decoder_input, need_pad=False).unsqueeze(1)
+        decoder_out = model.joiner.decoder_proj(decoder_out)
+        # decoder_out is of shape (num_hyps, 1, 1, joiner_dim)
+
+        # Note: For torch 1.7.1 and below, it requires a torch.int64 tensor
+        # as index, so we use `to(torch.int64)` below.
+        current_encoder_out = torch.index_select(
+            current_encoder_out,
+            dim=0,
+            index=hyps_shape.row_ids(1).to(torch.int64),
+        )  # (num_hyps, 1, 1, encoder_out_dim)
+
+        logits = model.joiner(
+            current_encoder_out, decoder_out, project_input=False
+        )  # (num_hyps, 1, 1, vocab_size)
+        logits = logits.squeeze(1).squeeze(1)  # (num_hyps, vocab_size)
+
+        log_probs = logits.log_softmax(dim=-1)  # (num_hyps, vocab_size)
+        log_probs.add_(ys_log_probs)
+
+        vocab_size = log_probs.size(-1)
+
+        row_splits = hyps_shape.row_splits(1) * vocab_size
+        log_probs_shape = k2.ragged.create_ragged_shape2(
+            row_splits=row_splits, cached_tot_size=log_probs.numel()
+        )
+        ragged_log_probs = k2.RaggedTensor(
+            shape=log_probs_shape, value=log_probs.reshape(-1)
+        )  # [batch][num_hyps*vocab_size]
+
+        for i in range(batch_size):
+            for h, hyp in enumerate(A[i]):
+                pos_u = len(hyp.timestamp)
+                idx_offset = h * vocab_size
+                if (sorted_encoder_out_lens[i] - 1 - t) >= (sorted_ys_lens[i] - pos_u):
+                    # emit blank token
+                    new_hyp = Hypothesis(
+                        log_prob=ragged_log_probs[i][idx_offset + blank_id],
+                        ys=hyp.ys[:],
+                        timestamp=hyp.timestamp[:],
+                    )
+                    B[i].add(new_hyp)
+                if pos_u < sorted_ys_lens[i]:
+                    # emit non-blank token
+                    new_token = sorted_ys_list[i][pos_u]
+                    new_hyp = Hypothesis(
+                        log_prob=ragged_log_probs[i][idx_offset + new_token],
+                        ys=hyp.ys + [new_token],
+                        timestamp=hyp.timestamp + [t],
+                    )
+                    B[i].add(new_hyp)
+
+            if len(B[i]) > beam_size:
+                B[i] = B[i].topk(beam_size, length_norm=True)
+
+    B = B + finalized_B
+    sorted_hyps = [b.get_most_probable() for b in B]
+    unsorted_indices = packed_encoder_out.unsorted_indices.tolist()
+    hyps = [sorted_hyps[i] for i in unsorted_indices]
+    ans = []
+    for i, hyp in enumerate(hyps):
+        assert hyp.ys[context_size:] == ys_list[i], (hyp.ys[context_size:], ys_list[i])
+        ans.append(hyp.timestamp)
+
+    return ans

egs/librispeech/ASR/pruned_transducer_stateless7/compute_ali.py

@@ -0,0 +1,345 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021-2023 Xiaomi Corporation (Author: Fangjun Kuang,
+#                                                 Zengwei Yao,
+#                                                 Xiaoyu Yang)
+#
+# 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.
+"""
+The script gets forced-alignments based on the modified_beam_search decoding method.
+Both token-level alignments and word-level alignments are saved to the new cuts manifests.
+
+It loads a checkpoint and uses it to get the forced-alignments.
+You can generate the checkpoint with the following command:
+
+./pruned_transducer_stateless7/export.py \
+  --exp-dir ./pruned_transducer_stateless7/exp \
+  --bpe-model data/lang_bpe_500/bpe.model \
+  --epoch 30 \
+  --avg 9
+
+Usage of this script:
+
+./pruned_transducer_stateless7/compute_ali.py \
+    --checkpoint ./pruned_transducer_stateless7/exp/pretrained.pt \
+    --bpe-model data/lang_bpe_500/bpe.model \
+    --dataset test-clean \
+    --max-duration 300 \
+    --beam-size 4 \
+    --cuts-out-dir data/fbank_ali_beam_search
+"""
+
+
+import argparse
+import logging
+from pathlib import Path
+from typing import List, Tuple
+
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from alignment import batch_force_alignment
+from asr_datamodule import LibriSpeechAsrDataModule
+from train import add_model_arguments, get_params, get_transducer_model
+
+from icefall.utils import AttributeDict, convert_timestamp, parse_timestamp
+from lhotse import CutSet
+from lhotse.serialization import SequentialJsonlWriter
+from lhotse.supervision import AlignmentItem
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--checkpoint",
+        type=str,
+        required=True,
+        help="Path to the checkpoint. "
+        "The checkpoint is assumed to be saved by "
+        "icefall.checkpoint.save_checkpoint().",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        required=True,
+        help="""The name of the dataset to compute alignments for.
+        Possible values are:
+        - test-clean
+        - test-other
+        - train-clean-100
+        - train-clean-360
+        - train-other-500
+        - dev-clean
+        - dev-other
+        """,
+    )
+
+    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(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    parser.add_argument(
+        "--cuts-out-dir",
+        type=str,
+        default="data/fbank_ali_beam_search",
+        help="The dir to save the new cuts manifests with alignments",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def align_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+) -> Tuple[List[List[str]], List[List[str]], List[List[float]], List[List[float]]]:
+    """Get forced-alignments for one 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`.
+
+    Returns:
+      token_list:
+        A list of token list.
+      word_list:
+        A list of word list.
+      token_time_list:
+        A list of timestamps list for tokens.
+      word_time_list.
+        A list of timestamps list for words.
+
+      where len(token_list) == len(word_list) == len(token_time_list) == len(word_time_list),
+      len(token_list[i]) == len(token_time_list[i]),
+      and len(word_list[i]) == len(word_time_list[i])
+
+    """
+    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)
+
+    encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
+
+    texts = supervisions["text"]
+    ys_list: List[List[int]] = sp.encode(texts, out_type=int)
+
+    frame_indexes = batch_force_alignment(
+        model, encoder_out, encoder_out_lens, ys_list, params.beam_size
+    )
+
+    token_list = []
+    word_list = []
+    token_time_list = []
+    word_time_list = []
+    for i in range(encoder_out.size(0)):
+        tokens = sp.id_to_piece(ys_list[i])
+        words = texts[i].split()
+        token_time = convert_timestamp(
+            frame_indexes[i], params.subsampling_factor, params.frame_shift_ms
+        )
+        word_time = parse_timestamp(tokens, token_time)
+        assert len(word_time) == len(words), (len(word_time), len(words))
+
+        token_list.append(tokens)
+        word_list.append(words)
+        token_time_list.append(token_time)
+        word_time_list.append(word_time)
+
+    return token_list, word_list, token_time_list, word_time_list
+
+
+def align_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    writer: SequentialJsonlWriter,
+) -> None:
+    """Get forced-alignments for the 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.
+      writer:
+        Writer to save the cuts with alignments.
+    """
+    log_interval = 20
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    for batch_idx, batch in enumerate(dl):
+        token_list, word_list, token_time_list, word_time_list = align_one_batch(
+            params=params, model=model, sp=sp, batch=batch
+        )
+
+        cut_list = batch["supervisions"]["cut"]
+        for cut, token, word, token_time, word_time in zip(
+            cut_list, token_list, word_list, token_time_list, word_time_list
+        ):
+            assert len(cut.supervisions) == 1, f"{len(cut.supervisions)}"
+            token_ali = [
+                AlignmentItem(
+                    symbol=token[i],
+                    start=round(token_time[i], ndigits=3),
+                    duration=None,
+                )
+                for i in range(len(token))
+            ]
+            word_ali = [
+                AlignmentItem(
+                    symbol=word[i], start=round(word_time[i], ndigits=3), duration=None
+                )
+                for i in range(len(word))
+            ]
+            cut.supervisions[0].alignment = {"word": word_ali, "token": token_ali}
+            writer.write(cut, flush=True)
+
+        num_cuts += len(cut_list)
+        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}")
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+
+    params = get_params()
+    params.update(vars(args))
+
+    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.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    model.load_state_dict(checkpoint["model"], strict=False)
+    model.to(device)
+    model.eval()
+
+    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
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    if params.dataset == "test-clean":
+        test_clean_cuts = librispeech.test_clean_cuts()
+        dl = librispeech.test_dataloaders(test_clean_cuts)
+    elif params.dataset == "test-other":
+        test_other_cuts = librispeech.test_other_cuts()
+        dl = librispeech.test_dataloaders(test_other_cuts)
+    elif params.dataset == "train-clean-100":
+        train_clean_100_cuts = librispeech.train_clean_100_cuts()
+        dl = librispeech.train_dataloaders(train_clean_100_cuts)
+    elif params.dataset == "train-clean-360":
+        train_clean_360_cuts = librispeech.train_clean_360_cuts()
+        dl = librispeech.train_dataloaders(train_clean_360_cuts)
+    elif params.dataset == "train-other-500":
+        train_other_500_cuts = librispeech.train_other_500_cuts()
+        dl = librispeech.train_dataloaders(train_other_500_cuts)
+    elif params.dataset == "dev-clean":
+        dev_clean_cuts = librispeech.dev_clean_cuts()
+        dl = librispeech.valid_dataloaders(dev_clean_cuts)
+    else:
+        assert params.dataset == "dev-other", f"{params.dataset}"
+        dev_other_cuts = librispeech.dev_other_cuts()
+        dl = librispeech.valid_dataloaders(dev_other_cuts)
+
+    cuts_out_dir = Path(params.cuts_out_dir)
+    cuts_out_dir.mkdir(parents=True, exist_ok=True)
+    cuts_out_path = cuts_out_dir / f"librispeech_cuts_{params.dataset}.jsonl.gz"
+
+    with CutSet.open_writer(cuts_out_path) as writer:
+        align_dataset(dl=dl, params=params, model=model, sp=sp, writer=writer)
+
+    logging.info(
+        f"For dataset {params.dataset}, the cut manifest with framewise token alignments "
+        f"and word alignments are saved to {cuts_out_path}"
+    )
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/librispeech/ASR/pruned_transducer_stateless7/decode.py

@@ -343,29 +343,6 @@ def get_parser():
         fast_beam_search_nbest_LG, and fast_beam_search_nbest_oracle""",
     )
 
-    parser.add_argument(
-        "--simulate-streaming",
-        type=str2bool,
-        default=False,
-        help="""Whether to simulate streaming in decoding, this is a good way to
-        test a streaming model.
-        """,
-    )
-
-    parser.add_argument(
-        "--decode-chunk-size",
-        type=int,
-        default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
-    )
-
-    parser.add_argument(
-        "--left-context",
-        type=int,
-        default=64,
-        help="left context can be seen during decoding (in frames after subsampling)",
-    )
-
     parser.add_argument(
         "--use-shallow-fusion",
         type=str2bool,
@@ -474,22 +451,7 @@ def decode_one_batch(
     supervisions = batch["supervisions"]
     feature_lens = supervisions["num_frames"].to(device)
 
-    if params.simulate_streaming:
+    encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
-        feature_lens += params.left_context
-        feature = torch.nn.functional.pad(
-            feature,
-            pad=(0, 0, 0, params.left_context),
-            value=LOG_EPS,
-        )
-        encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
-            x=feature,
-            x_lens=feature_lens,
-            chunk_size=params.decode_chunk_size,
-            left_context=params.left_context,
-            simulate_streaming=True,
-        )
-    else:
-        encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
 
     hyps = []
 
@@ -571,7 +533,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LM=LM,
         )
         for hyp in sp.decode(hyp_tokens):
@@ -582,7 +543,6 @@ def decode_one_batch(
             encoder_out=encoder_out,
             encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
-            sp=sp,
             LODR_lm=ngram_lm,
             LODR_lm_scale=ngram_lm_scale,
             LM=LM,
@@ -782,10 +742,6 @@ def main():
     else:
         params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
 
-    if params.simulate_streaming:
-        params.suffix += f"-streaming-chunk-size-{params.decode_chunk_size}"
-        params.suffix += f"-left-context-{params.left_context}"
-
     if "fast_beam_search" in params.decoding_method:
         params.suffix += f"-beam-{params.beam}"
         params.suffix += f"-max-contexts-{params.max_contexts}"
@@ -834,11 +790,6 @@ def main():
     params.unk_id = sp.piece_to_id("<unk>")
     params.vocab_size = sp.get_piece_size()
 
-    if params.simulate_streaming:
-        assert (
-            params.causal_convolution
-        ), "Decoding in streaming requires causal convolution"
-
     logging.info(params)
 
     logging.info("About to create model")

egs/librispeech/ASR/pruned_transducer_stateless7/export-onnx.py

@@ -55,6 +55,7 @@ import sentencepiece as spm
 import torch
 import torch.nn as nn
 from decoder import Decoder
+from onnxruntime.quantization import QuantType, quantize_dynamic
 from scaling_converter import convert_scaled_to_non_scaled
 from train import add_model_arguments, get_params, get_transducer_model
 from zipformer import Zipformer
@@ -291,6 +292,16 @@ def export_encoder_model_onnx(
         },
     )
 
+    meta_data = {
+        "model_type": "zipformer",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "comment": "stateless7",
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
 
 def export_decoder_model_onnx(
     decoder_model: OnnxDecoder,
@@ -553,6 +564,35 @@ def main():
     )
     logging.info(f"Exported joiner to {joiner_filename}")
 
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
 
 if __name__ == "__main__":
     formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"

egs/librispeech/ASR/pruned_transducer_stateless7/test_compute_ali.py

@@ -0,0 +1,130 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021-2023 Xiaomi Corporation (Author: Fangjun Kuang,
+#                                                 Zengwei Yao)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This script compares the word-level alignments generated based on modified_beam_search decoding
+(in ./pruned_transducer_stateless7/compute_ali.py) to the reference alignments generated
+by torchaudio framework (in ./add_alignments.sh).
+
+Usage:
+
+./pruned_transducer_stateless7/compute_ali.py \
+    --checkpoint ./pruned_transducer_stateless7/exp/pretrained.pt \
+    --bpe-model data/lang_bpe_500/bpe.model \
+    --dataset test-clean \
+    --max-duration 300 \
+    --beam-size 4 \
+    --cuts-out-dir data/fbank_ali_beam_search
+
+And the you can run:
+
+./pruned_transducer_stateless7/test_compute_ali.py \
+  --cuts-out-dir ./data/fbank_ali_test \
+  --cuts-ref-dir ./data/fbank_ali_torch \
+  --dataset train-clean-100
+"""
+import argparse
+import logging
+from pathlib import Path
+
+import torch
+from lhotse import load_manifest
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--cuts-out-dir",
+        type=Path,
+        default="./data/fbank_ali",
+        help="The dir that saves the generated cuts manifests with alignments",
+    )
+
+    parser.add_argument(
+        "--cuts-ref-dir",
+        type=Path,
+        default="./data/fbank_ali_torch",
+        help="The dir that saves the reference cuts manifests with alignments",
+    )
+
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        required=True,
+        help="""The name of the dataset:
+        Possible values are:
+            - test-clean
+            - test-other
+            - train-clean-100
+            - train-clean-360
+            - train-other-500
+            - dev-clean
+            - dev-other
+        """,
+    )
+
+    return parser
+
+
+def main():
+    args = get_parser().parse_args()
+
+    cuts_out_jsonl = args.cuts_out_dir / f"librispeech_cuts_{args.dataset}.jsonl.gz"
+    cuts_ref_jsonl = args.cuts_ref_dir / f"librispeech_cuts_{args.dataset}.jsonl.gz"
+
+    logging.info(f"Loading {cuts_out_jsonl} and {cuts_ref_jsonl}")
+    cuts_out = load_manifest(cuts_out_jsonl)
+    cuts_ref = load_manifest(cuts_ref_jsonl)
+    cuts_ref = cuts_ref.sort_like(cuts_out)
+
+    all_time_diffs = []
+    for cut_out, cut_ref in zip(cuts_out, cuts_ref):
+        time_out = [
+            ali.start
+            for ali in cut_out.supervisions[0].alignment["word"]
+            if ali.symbol != ""
+        ]
+        time_ref = [
+            ali.start
+            for ali in cut_ref.supervisions[0].alignment["word"]
+            if ali.symbol != ""
+        ]
+        assert len(time_out) == len(time_ref), (len(time_out), len(time_ref))
+        diff = [
+            round(abs(out - ref), ndigits=3) for out, ref in zip(time_out, time_ref)
+        ]
+        all_time_diffs += diff
+
+    all_time_diffs = torch.tensor(all_time_diffs)
+    logging.info(
+        f"For the word-level alignments abs difference on dataset {args.dataset}, "
+        f"mean: {'%.2f' % all_time_diffs.mean()}s, std: {'%.2f' % all_time_diffs.std()}s"
+    )
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/export-onnx-zh.py

@@ -0,0 +1,678 @@
+#!/usr/bin/env python3
+#
+# Copyright 2023 Xiaomi Corporation (Author: Fangjun Kuang)
+
+"""
+This script exports a transducer model from PyTorch to ONNX.
+
+We use the pre-trained model from
+https://huggingface.co/pfluo/k2fsa-zipformer-chinese-english-mixed
+as an example to show how to use this file.
+
+1. Download the pre-trained model
+
+cd egs/librispeech/ASR
+
+repo_url=https://huggingface.co/pfluo/k2fsa-zipformer-chinese-english-mixed
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+repo=$(basename $repo_url)
+
+pushd $repo
+git lfs pull --include "data/lang_char_bpe/L.pt"
+git lfs pull --include "data/lang_char_bpe/Linv.pt"
+git lfs pull --include "data/lang_char_bpe/L_disambig.pt"
+git lfs pull --include "exp/pretrained.pt"
+cd exp
+ln -s pretrained.pt epoch-99.pt
+popd
+
+2. Export the model to ONNX
+
+./pruned_transducer_stateless7_streaming/export-onnx-zh.py \
+  --lang-dir $repo/data/lang_char_bpe \
+  --use-averaged-model 0 \
+  --epoch 99 \
+  --avg 1 \
+  --exp-dir $repo/exp/ \
+  --decode-chunk-len 32 \
+  --num-encoder-layers "2,4,3,2,4" \
+  --feedforward-dims "1024,1024,1536,1536,1024" \
+  --nhead "8,8,8,8,8" \
+  --encoder-dims "384,384,384,384,384" \
+  --attention-dims "192,192,192,192,192" \
+  --encoder-unmasked-dims "256,256,256,256,256" \
+  --zipformer-downsampling-factors "1,2,4,8,2" \
+  --cnn-module-kernels "31,31,31,31,31" \
+  --decoder-dim 512 \
+  --joiner-dim 512
+
+It will generate the following 3 files in $repo/exp
+
+  - encoder-epoch-99-avg-1.onnx
+  - decoder-epoch-99-avg-1.onnx
+  - joiner-epoch-99-avg-1.onnx
+
+See ./onnx_pretrained.py for how to use the exported models.
+"""
+
+import argparse
+import logging
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import onnx
+import torch
+import torch.nn as nn
+from decoder import Decoder
+from onnxruntime.quantization import QuantType, quantize_dynamic
+from scaling_converter import convert_scaled_to_non_scaled
+from torch import Tensor
+from train import add_model_arguments, get_params, get_transducer_model
+from zipformer import Zipformer
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.lexicon import Lexicon
+from icefall.utils import setup_logger, str2bool
+
+
+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="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_char",
+        help="The lang dir",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+class OnnxEncoder(nn.Module):
+    """A wrapper for Zipformer and the encoder_proj from the joiner"""
+
+    def __init__(self, encoder: Zipformer, encoder_proj: nn.Linear):
+        """
+        Args:
+          encoder:
+            A Zipformer encoder.
+          encoder_proj:
+            The projection layer for encoder from the joiner.
+        """
+        super().__init__()
+        self.encoder = encoder
+        self.encoder_proj = encoder_proj
+
+    def forward(self, x: Tensor, states: List[Tensor]) -> Tuple[Tensor, List[Tensor]]:
+        """Please see the help information of Zipformer.streaming_forward"""
+        N = x.size(0)
+        T = x.size(1)
+        x_lens = torch.tensor([T] * N, device=x.device)
+
+        output, _, new_states = self.encoder.streaming_forward(
+            x=x,
+            x_lens=x_lens,
+            states=states,
+        )
+
+        output = self.encoder_proj(output)
+        # Now output is of shape (N, T, joiner_dim)
+
+        return output, new_states
+
+
+class OnnxDecoder(nn.Module):
+    """A wrapper for Decoder and the decoder_proj from the joiner"""
+
+    def __init__(self, decoder: Decoder, decoder_proj: nn.Linear):
+        super().__init__()
+        self.decoder = decoder
+        self.decoder_proj = decoder_proj
+
+    def forward(self, y: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, context_size).
+        Returns
+          Return a 2-D tensor of shape (N, joiner_dim)
+        """
+        need_pad = False
+        decoder_output = self.decoder(y, need_pad=need_pad)
+        decoder_output = decoder_output.squeeze(1)
+        output = self.decoder_proj(decoder_output)
+
+        return output
+
+
+class OnnxJoiner(nn.Module):
+    """A wrapper for the joiner"""
+
+    def __init__(self, output_linear: nn.Linear):
+        super().__init__()
+        self.output_linear = output_linear
+
+    def forward(
+        self,
+        encoder_out: torch.Tensor,
+        decoder_out: torch.Tensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+          decoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+        Returns:
+          Return a 2-D tensor of shape (N, vocab_size)
+        """
+        logit = encoder_out + decoder_out
+        logit = self.output_linear(torch.tanh(logit))
+        return logit
+
+
+def add_meta_data(filename: str, meta_data: Dict[str, str]):
+    """Add meta data to an ONNX model. It is changed in-place.
+
+    Args:
+      filename:
+        Filename of the ONNX model to be changed.
+      meta_data:
+        Key-value pairs.
+    """
+    model = onnx.load(filename)
+    for key, value in meta_data.items():
+        meta = model.metadata_props.add()
+        meta.key = key
+        meta.value = value
+
+    onnx.save(model, filename)
+
+
+def export_encoder_model_onnx(
+    encoder_model: OnnxEncoder,
+    encoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """
+    Onnx model inputs:
+      - 0: src
+      - many state tensors (the exact number depending on the actual model)
+
+    Onnx model outputs:
+      - 0: output, its shape is (N, T, joiner_dim)
+      - many state tensors (the exact number depending on the actual model)
+
+    Args:
+      encoder_model:
+        The model to be exported
+      encoder_filename:
+        The filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+
+    encoder_model.encoder.__class__.forward = (
+        encoder_model.encoder.__class__.streaming_forward
+    )
+
+    decode_chunk_len = encoder_model.encoder.decode_chunk_size * 2
+    pad_length = 7
+    T = decode_chunk_len + pad_length
+    logging.info(f"decode_chunk_len: {decode_chunk_len}")
+    logging.info(f"pad_length: {pad_length}")
+    logging.info(f"T: {T}")
+
+    x = torch.rand(1, T, 80, dtype=torch.float32)
+
+    init_state = encoder_model.encoder.get_init_state()
+
+    num_encoders = encoder_model.encoder.num_encoders
+    logging.info(f"num_encoders: {num_encoders}")
+    logging.info(f"len(init_state): {len(init_state)}")
+
+    inputs = {}
+    input_names = ["x"]
+
+    outputs = {}
+    output_names = ["encoder_out"]
+
+    def build_inputs_outputs(tensors, name, N):
+        for i, s in enumerate(tensors):
+            logging.info(f"{name}_{i}.shape: {s.shape}")
+            inputs[f"{name}_{i}"] = {N: "N"}
+            outputs[f"new_{name}_{i}"] = {N: "N"}
+            input_names.append(f"{name}_{i}")
+            output_names.append(f"new_{name}_{i}")
+
+    num_encoder_layers = ",".join(map(str, encoder_model.encoder.num_encoder_layers))
+    encoder_dims = ",".join(map(str, encoder_model.encoder.encoder_dims))
+    attention_dims = ",".join(map(str, encoder_model.encoder.attention_dims))
+    cnn_module_kernels = ",".join(map(str, encoder_model.encoder.cnn_module_kernels))
+    ds = encoder_model.encoder.zipformer_downsampling_factors
+    left_context_len = encoder_model.encoder.left_context_len
+    left_context_len = [left_context_len // k for k in ds]
+    left_context_len = ",".join(map(str, left_context_len))
+
+    meta_data = {
+        "model_type": "zipformer",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "decode_chunk_len": str(decode_chunk_len),  # 32
+        "T": str(T),  # 39
+        "num_encoder_layers": num_encoder_layers,
+        "encoder_dims": encoder_dims,
+        "attention_dims": attention_dims,
+        "cnn_module_kernels": cnn_module_kernels,
+        "left_context_len": left_context_len,
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    # (num_encoder_layers, 1)
+    cached_len = init_state[num_encoders * 0 : num_encoders * 1]
+
+    # (num_encoder_layers, 1, encoder_dim)
+    cached_avg = init_state[num_encoders * 1 : num_encoders * 2]
+
+    # (num_encoder_layers, left_context_len, 1, attention_dim)
+    cached_key = init_state[num_encoders * 2 : num_encoders * 3]
+
+    # (num_encoder_layers, left_context_len, 1, attention_dim//2)
+    cached_val = init_state[num_encoders * 3 : num_encoders * 4]
+
+    # (num_encoder_layers, left_context_len, 1, attention_dim//2)
+    cached_val2 = init_state[num_encoders * 4 : num_encoders * 5]
+
+    # (num_encoder_layers, 1, encoder_dim, cnn_module_kernel-1)
+    cached_conv1 = init_state[num_encoders * 5 : num_encoders * 6]
+
+    # (num_encoder_layers, 1, encoder_dim, cnn_module_kernel-1)
+    cached_conv2 = init_state[num_encoders * 6 : num_encoders * 7]
+
+    build_inputs_outputs(cached_len, "cached_len", 1)
+    build_inputs_outputs(cached_avg, "cached_avg", 1)
+    build_inputs_outputs(cached_key, "cached_key", 2)
+    build_inputs_outputs(cached_val, "cached_val", 2)
+    build_inputs_outputs(cached_val2, "cached_val2", 2)
+    build_inputs_outputs(cached_conv1, "cached_conv1", 1)
+    build_inputs_outputs(cached_conv2, "cached_conv2", 1)
+
+    logging.info(inputs)
+    logging.info(outputs)
+    logging.info(input_names)
+    logging.info(output_names)
+
+    torch.onnx.export(
+        encoder_model,
+        (x, init_state),
+        encoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=input_names,
+        output_names=output_names,
+        dynamic_axes={
+            "x": {0: "N"},
+            "encoder_out": {0: "N"},
+            **inputs,
+            **outputs,
+        },
+    )
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
+
+def export_decoder_model_onnx(
+    decoder_model: nn.Module,
+    decoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the decoder model to ONNX format.
+
+    The exported model has one input:
+
+        - y: a torch.int64 tensor of shape (N, context_size)
+
+    and has one output:
+
+        - decoder_out: a torch.float32 tensor of shape (N, joiner_dim)
+
+    Note: The argument need_pad is fixed to False.
+
+    Args:
+      decoder_model:
+        The decoder model to be exported.
+      decoder_filename:
+        Filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+    context_size = decoder_model.decoder.context_size
+    vocab_size = decoder_model.decoder.vocab_size
+    y = torch.zeros(10, context_size, dtype=torch.int64)
+    torch.onnx.export(
+        decoder_model,
+        y,
+        decoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=["y"],
+        output_names=["decoder_out"],
+        dynamic_axes={
+            "y": {0: "N"},
+            "decoder_out": {0: "N"},
+        },
+    )
+    meta_data = {
+        "context_size": str(context_size),
+        "vocab_size": str(vocab_size),
+    }
+    add_meta_data(filename=decoder_filename, meta_data=meta_data)
+
+
+def export_joiner_model_onnx(
+    joiner_model: nn.Module,
+    joiner_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the joiner model to ONNX format.
+    The exported joiner model has two inputs:
+
+        - encoder_out: a tensor of shape (N, joiner_dim)
+        - decoder_out: a tensor of shape (N, joiner_dim)
+
+    and produces one output:
+
+        - logit: a tensor of shape (N, vocab_size)
+    """
+    joiner_dim = joiner_model.output_linear.weight.shape[1]
+    logging.info(f"joiner dim: {joiner_dim}")
+
+    projected_encoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+    projected_decoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+
+    torch.onnx.export(
+        joiner_model,
+        (projected_encoder_out, projected_decoder_out),
+        joiner_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=[
+            "encoder_out",
+            "decoder_out",
+        ],
+        output_names=["logit"],
+        dynamic_axes={
+            "encoder_out": {0: "N"},
+            "decoder_out": {0: "N"},
+            "logit": {0: "N"},
+        },
+    )
+    meta_data = {
+        "joiner_dim": str(joiner_dim),
+    }
+    add_meta_data(filename=joiner_filename, meta_data=meta_data)
+
+
+@torch.no_grad()
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    setup_logger(f"{params.exp_dir}/log-export/log-export-onnx")
+
+    logging.info(f"device: {device}")
+
+    lexicon = Lexicon(params.lang_dir)
+    params.blank_id = 0
+    params.vocab_size = max(lexicon.tokens) + 1
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    model.to(device)
+
+    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("cpu")
+    model.eval()
+
+    convert_scaled_to_non_scaled(model, inplace=True)
+    encoder = OnnxEncoder(
+        encoder=model.encoder,
+        encoder_proj=model.joiner.encoder_proj,
+    )
+
+    decoder = OnnxDecoder(
+        decoder=model.decoder,
+        decoder_proj=model.joiner.decoder_proj,
+    )
+
+    joiner = OnnxJoiner(output_linear=model.joiner.output_linear)
+
+    encoder_num_param = sum([p.numel() for p in encoder.parameters()])
+    decoder_num_param = sum([p.numel() for p in decoder.parameters()])
+    joiner_num_param = sum([p.numel() for p in joiner.parameters()])
+    total_num_param = encoder_num_param + decoder_num_param + joiner_num_param
+    logging.info(f"encoder parameters: {encoder_num_param}")
+    logging.info(f"decoder parameters: {decoder_num_param}")
+    logging.info(f"joiner parameters: {joiner_num_param}")
+    logging.info(f"total parameters: {total_num_param}")
+
+    if params.iter > 0:
+        suffix = f"iter-{params.iter}"
+    else:
+        suffix = f"epoch-{params.epoch}"
+
+    suffix += f"-avg-{params.avg}"
+    if params.use_averaged_model:
+        suffix += "-with-averaged-model"
+
+    opset_version = 13
+
+    logging.info("Exporting encoder")
+    encoder_filename = params.exp_dir / f"encoder-{suffix}.onnx"
+    export_encoder_model_onnx(
+        encoder,
+        encoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported encoder to {encoder_filename}")
+
+    logging.info("Exporting decoder")
+    decoder_filename = params.exp_dir / f"decoder-{suffix}.onnx"
+    export_decoder_model_onnx(
+        decoder,
+        decoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported decoder to {decoder_filename}")
+
+    logging.info("Exporting joiner")
+    joiner_filename = params.exp_dir / f"joiner-{suffix}.onnx"
+    export_joiner_model_onnx(
+        joiner,
+        joiner_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported joiner to {joiner_filename}")
+
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+
+if __name__ == "__main__":
+    main()

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

@@ -53,6 +53,7 @@ import sentencepiece as spm
 import torch
 import torch.nn as nn
 from decoder import Decoder
+from onnxruntime.quantization import QuantType, quantize_dynamic
 from scaling_converter import convert_scaled_to_non_scaled
 from torch import Tensor
 from train import add_model_arguments, get_params, get_transducer_model
@@ -634,6 +635,35 @@ def main():
     )
     logging.info(f"Exported joiner to {joiner_filename}")
 
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
 
 if __name__ == "__main__":
     main()

egs/librispeech/ASR/pruned_transducer_stateless8/decode.py

@@ -301,29 +301,6 @@ def get_parser():
         fast_beam_search_nbest_LG, and fast_beam_search_nbest_oracle""",
     )
 
-    parser.add_argument(
-        "--simulate-streaming",
-        type=str2bool,
-        default=False,
-        help="""Whether to simulate streaming in decoding, this is a good way to
-        test a streaming model.
-        """,
-    )
-
-    parser.add_argument(
-        "--decode-chunk-size",
-        type=int,
-        default=16,
-        help="The chunk size for decoding (in frames after subsampling)",
-    )
-
-    parser.add_argument(
-        "--left-context",
-        type=int,
-        default=64,
-        help="left context can be seen during decoding (in frames after subsampling)",
-    )
-
     add_model_arguments(parser)
 
     return parser
@@ -378,22 +355,7 @@ def decode_one_batch(
     supervisions = batch["supervisions"]
     feature_lens = supervisions["num_frames"].to(device)
 
-    if params.simulate_streaming:
+    encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
-        feature_lens += params.left_context
-        feature = torch.nn.functional.pad(
-            feature,
-            pad=(0, 0, 0, params.left_context),
-            value=LOG_EPS,
-        )
-        encoder_out, encoder_out_lens, _ = model.encoder.streaming_forward(
-            x=feature,
-            x_lens=feature_lens,
-            chunk_size=params.decode_chunk_size,
-            left_context=params.left_context,
-            simulate_streaming=True,
-        )
-    else:
-        encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
 
     hyps = []
 
@@ -651,10 +613,6 @@ def main():
     else:
         params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
 
-    if params.simulate_streaming:
-        params.suffix += f"-streaming-chunk-size-{params.decode_chunk_size}"
-        params.suffix += f"-left-context-{params.left_context}"
-
     if "fast_beam_search" in params.decoding_method:
         params.suffix += f"-beam-{params.beam}"
         params.suffix += f"-max-contexts-{params.max_contexts}"
@@ -690,11 +648,6 @@ def main():
     params.unk_id = sp.piece_to_id("<unk>")
     params.vocab_size = sp.get_piece_size()
 
-    if params.simulate_streaming:
-        assert (
-            params.causal_convolution
-        ), "Decoding in streaming requires causal convolution"
-
     logging.info(params)
 
     logging.info("About to create model")

egs/librispeech/ASR/transducer_stateless/conformer.py

@@ -358,6 +358,11 @@ class Conformer(Transformer):
 
             assert x.size(0) == lengths.max().item()
 
+            if chunk_size < 0:
+                # use full attention
+                chunk_size = x.size(0)
+                left_context = -1
+
             num_left_chunks = -1
             if left_context >= 0:
                 assert left_context % chunk_size == 0

egs/wenetspeech/ASR/finetune.sh

@@ -0,0 +1,82 @@
+#!/usr/bin/env bash
+
+# fix segmentation fault reported in https://github.com/k2-fsa/icefall/issues/674
+export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
+
+set -eou pipefail
+
+stage=-1
+stop_stage=100
+
+# This is an example script for fine-tuning. Here, we fine-tune a model trained
+# on WenetSpeech on Aishell. The model used for fine-tuning is
+# pruned_transducer_stateless2 (zipformer). If you want to fine-tune model
+# from another recipe, you can adapt ./pruned_transducer_stateless2/finetune.py
+# for that recipe. If you have any problem, please open up an issue in https://github.com/k2-fsa/icefall/issues.
+
+# We assume that you have already prepared the Aishell manfiest&features under ./data.
+# If you haven't done that, please see https://github.com/k2-fsa/icefall/blob/master/egs/aishell/ASR/prepare.sh.
+
+. shared/parse_options.sh || exit 1
+
+log() {
+  # This function is from espnet
+  local fname=${BASH_SOURCE[1]##*/}
+  echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+if [ $stage -le -1 ] && [ $stop_stage -ge -1 ]; then
+  log "Stage -1: Download Pre-trained model"
+
+  # clone from huggingface
+  git lfs install
+  git clone https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
+
+fi
+
+if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
+  log "Stage 0: Start fine-tuning"
+
+  # The following configuration of lr schedule should work well
+  # You may also tune the following parameters to adjust learning rate schedule
+  initial_lr=0.0001
+  lr_epochs=100
+  lr_batches=100000
+
+  # We recommend to start from an averaged model
+  finetune_ckpt=icefall_asr_wenetspeech_pruned_transducer_stateless2/exp/pretrained_epoch_10_avg_2.pt
+  lang_dir=icefall_asr_wenetspeech_pruned_transducer_stateless2/data/lang_char
+  export CUDA_VISIBLE_DEVICES="0,1"
+
+  ./pruned_transducer_stateless2/finetune.py \
+    --world-size 2 \
+    --master-port 18180 \
+    --num-epochs 15 \
+    --context-size 2 \
+    --exp-dir pruned_transducer_stateless2/exp_aishell_finetune \
+    --initial-lr $initial_lr \
+    --lr-epochs $lr_epochs \
+    --lr-batches $lr_batches \
+    --lang-dir $lang_dir \
+    --do-finetune True \
+    --finetune-ckpt $finetune_ckpt \
+    --max-duration 200
+fi
+
+if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
+  log "Stage 1: Decoding"
+
+  epoch=4
+  avg=4
+
+  for m in greedy_search modified_beam_search; do
+    python pruned_transducer_stateless2/decode_aishell.py \
+    --epoch $epoch \
+    --avg $avg \
+    --context-size 2 \
+    --beam-size 4 \
+    --exp-dir pruned_transducer_stateless2/exp_aishell_finetune \
+    --max-duration 400 \
+    --decoding-method $m
+  done
+fi

egs/wenetspeech/ASR/local/text2segments.py

@@ -40,8 +40,8 @@ from tqdm import tqdm
 # and 'data()' is only supported in static graph mode. So if you
 # want to use this api, should call 'paddle.enable_static()' before
 # this api to enter static graph mode.
-paddle.enable_static()
+# paddle.enable_static()
-paddle.disable_signal_handler()
+# paddle.disable_signal_handler()
 jieba.enable_paddle()
 
 

egs/wenetspeech/ASR/prepare.sh

@@ -261,3 +261,107 @@ if [ $stage -le 18 ] && [ $stop_stage -ge 18 ]; then
   log "Stage 18: Compile LG"
   python ./local/compile_lg.py --lang-dir $lang_char_dir
 fi
+
+# prepare RNNLM data
+if [ $stage -le 19 ] && [ $stop_stage -ge 19 ]; then
+  log "Stage 19: Prepare LM training data"
+
+  log "Processing char based data"
+  text_out_dir=data/lm_char
+
+  mkdir -p $text_out_dir
+
+  log "Genearating training text data"
+  
+  if [ ! -f $text_out_dir/lm_data.pt ]; then
+    ./local/prepare_char_lm_training_data.py \
+      --lang-char data/lang_char \
+      --lm-data $lang_char_dir/text_words_segmentation \
+      --lm-archive $text_out_dir/lm_data.pt
+  fi
+
+  log "Generating DEV text data"
+  # prepare validation text data 
+  if [ ! -f $text_out_dir/valid_text_words_segmentation ]; then
+    valid_text=${text_out_dir}/
+
+    gunzip -c data/manifests/wenetspeech_supervisions_DEV.jsonl.gz \
+      | jq '.text' | sed 's/"//g' \
+      | ./local/text2token.py -t "char" > $text_out_dir/valid_text
+    
+    python3 ./local/text2segments.py \
+      --num-process $nj \
+      --input-file $text_out_dir/valid_text \
+      --output-file $text_out_dir/valid_text_words_segmentation
+  fi
+
+  ./local/prepare_char_lm_training_data.py \
+    --lang-char data/lang_char \
+    --lm-data $text_out_dir/valid_text_words_segmentation \
+    --lm-archive $text_out_dir/lm_data_valid.pt
+
+  # prepare TEST text data 
+  if [ ! -f $text_out_dir/TEST_text_words_segmentation ]; then
+    log "Prepare text for test set."
+    for test_set in TEST_MEETING TEST_NET; do
+        gunzip -c data/manifests/wenetspeech_supervisions_${test_set}.jsonl.gz \
+          | jq '.text' | sed 's/"//g' \
+          | ./local/text2token.py -t "char" > $text_out_dir/${test_set}_text
+
+        python3 ./local/text2segments.py \
+          --num-process $nj \
+          --input-file $text_out_dir/${test_set}_text \
+          --output-file $text_out_dir/${test_set}_text_words_segmentation
+    done
+    
+    cat $text_out_dir/TEST_*_text_words_segmentation > $text_out_dir/test_text_words_segmentation
+  fi
+
+  ./local/prepare_char_lm_training_data.py \
+    --lang-char data/lang_char \
+    --lm-data $text_out_dir/test_text_words_segmentation \
+    --lm-archive $text_out_dir/lm_data_test.pt
+
+fi
+
+# sort RNNLM data
+if [ $stage -le 20 ] && [ $stop_stage -ge 20 ]; then
+  text_out_dir=data/lm_char
+
+  log "Sort lm data"
+
+  ./local/sort_lm_training_data.py \
+    --in-lm-data $text_out_dir/lm_data.pt \
+    --out-lm-data $text_out_dir/sorted_lm_data.pt \
+    --out-statistics $text_out_dir/statistics.txt
+
+  ./local/sort_lm_training_data.py \
+    --in-lm-data $text_out_dir/lm_data_valid.pt \
+    --out-lm-data $text_out_dir/sorted_lm_data-valid.pt \
+    --out-statistics $text_out_dir/statistics-valid.txt
+
+  ./local/sort_lm_training_data.py \
+    --in-lm-data $text_out_dir/lm_data_test.pt \
+    --out-lm-data $text_out_dir/sorted_lm_data-test.pt \
+    --out-statistics $text_out_dir/statistics-test.txt
+fi
+
+export CUDA_VISIBLE_DEVICES="0,1"
+
+if [ $stage -le 21 ] && [ $stop_stage -ge 21 ]; then
+  log "Stage 21: Train RNN LM model"
+  python ../../../icefall/rnn_lm/train.py \
+    --start-epoch 0 \
+    --world-size 2 \
+    --num-epochs 20 \
+    --use-fp16 0 \
+    --embedding-dim 2048 \
+    --hidden-dim 2048 \
+    --num-layers 2 \
+    --batch-size 400 \
+    --exp-dir rnnlm_char/exp \
+    --lm-data data/lm_char/sorted_lm_data.pt \
+    --lm-data-valid data/lm_char/sorted_lm_data-valid.pt \
+    --vocab-size 4336 \
+    --master-port 12340
+fi

egs/wenetspeech/ASR/pruned_transducer_stateless2/aishell.py

@@ -0,0 +1 @@
+../../../aishell/ASR/tdnn_lstm_ctc/asr_datamodule.py

egs/wenetspeech/ASR/pruned_transducer_stateless2/decode_aishell.py

@@ -0,0 +1,547 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021-2022 Xiaomi Corporation (Author: Fangjun Kuang,
+#                                                 Zengwei Yao)
+#
+# 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_stateless2/decode.py \
+    --epoch 84 \
+    --avg 25 \
+    --exp-dir ./pruned_transducer_stateless2/exp \
+    --max-duration 600 \
+    --decoding-method greedy_search
+
+(2) beam search (not recommended)
+./pruned_transducer_stateless2/decode.py \
+    --epoch 84 \
+    --avg 25 \
+    --exp-dir ./pruned_transducer_stateless2/exp \
+    --max-duration 600 \
+    --decoding-method beam_search \
+    --beam-size 4
+
+(3) modified beam search
+./pruned_transducer_stateless2/decode.py \
+    --epoch 84 \
+    --avg 25 \
+    --exp-dir ./pruned_transducer_stateless2/exp \
+    --max-duration 600 \
+    --decoding-method modified_beam_search \
+    --beam-size 4
+
+(4) fast beam search
+./pruned_transducer_stateless2/decode.py \
+    --epoch 84 \
+    --avg 25 \
+    --exp-dir ./pruned_transducer_stateless2/exp \
+    --max-duration 600 \
+    --decoding-method fast_beam_search \
+    --beam 4 \
+    --max-contexts 4 \
+    --max-states 8
+"""
+
+
+import argparse
+import logging
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import torch
+import torch.nn as nn
+from aishell import AishellAsrDataModule
+from beam_search import (
+    beam_search,
+    fast_beam_search_one_best,
+    greedy_search,
+    greedy_search_batch,
+    modified_beam_search,
+)
+from finetune import get_params, get_transducer_model
+
+from icefall.checkpoint import average_checkpoints, find_checkpoints, load_checkpoint
+from icefall.lexicon import Lexicon
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    write_error_stats,
+)
+
+
+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=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="pruned_transducer_stateless2/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=str,
+        default="data/lang_char",
+        help="The lang dir",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+          - modified_beam_search
+          - fast_beam_search
+        """,
+    )
+
+    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=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=8,
+        help="""Used only when --decoding-method is
+        fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=1,
+        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""",
+    )
+
+    return parser
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    token_table: k2.SymbolTable,
+    batch: dict,
+    decoding_graph: Optional[k2.Fsa] = None,
+) -> Dict[str, List[List[str]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+
+        - key: It indicates the setting used for decoding. For example,
+               if 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.
+      token_table:
+        It maps token ID to a string.
+      batch:
+        It is the return value from iterating
+        `lhotse.dataset.K2SpeechRecognitionDataset`. See its documentation
+        for the format of the `batch`.
+      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 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)
+
+    encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
+
+    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,
+        )
+    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,
+        )
+    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,
+        )
+    else:
+        hyp_tokens = []
+        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}"
+                )
+            hyp_tokens.append(hyp)
+
+    hyps = [[token_table[t] for t in tokens] for tokens in hyp_tokens]
+
+    if params.decoding_method == "greedy_search":
+        return {"greedy_search": hyps}
+    elif params.decoding_method == "fast_beam_search":
+        return {
+            (
+                f"beam_{params.beam}_"
+                f"max_contexts_{params.max_contexts}_"
+                f"max_states_{params.max_states}"
+            ): hyps
+        }
+    else:
+        return {f"beam_size_{params.beam_size}": hyps}
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    token_table: k2.SymbolTable,
+    decoding_graph: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      token_table:
+        It maps a token ID to a string.
+      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.
+    """
+    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,
+            token_table=token_table,
+            decoding_graph=decoding_graph,
+            batch=batch,
+        )
+
+        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_wers = 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 WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
+        # we compute CER for aishell dataset.
+        results_char = []
+        for res in results:
+            results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results_char, enable_log=True
+            )
+            test_set_wers[key] = wer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = params.res_dir / f"wer-summary-{test_set_name}-{params.suffix}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    AishellAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+    args.lang_dir = Path(args.lang_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    assert params.decoding_method in (
+        "greedy_search",
+        "beam_search",
+        "fast_beam_search",
+        "modified_beam_search",
+    )
+    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}"
+
+    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}"
+    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}"
+
+    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}")
+
+    lexicon = Lexicon(params.lang_dir)
+    params.blank_id = 0
+    params.vocab_size = max(lexicon.tokens) + 1
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    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 --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), strict=False
+        )
+    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), strict=False
+        )
+
+    model.to(device)
+    model.eval()
+
+    if params.decoding_method == "fast_beam_search":
+        decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)
+    else:
+        decoding_graph = None
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    aishell = AishellAsrDataModule(args)
+    test_cuts = aishell.test_cuts()
+    dev_cuts = aishell.valid_cuts()
+    test_dl = aishell.test_dataloaders(test_cuts)
+    dev_dl = aishell.test_dataloaders(dev_cuts)
+
+    test_sets = ["test", "dev"]
+    test_dls = [test_dl, dev_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dls):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            token_table=lexicon.token_table,
+            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/wenetspeech/ASR/pruned_transducer_stateless5/decode.py

@@ -2,6 +2,7 @@
 #
 # Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
 # Copyright 2022 Xiaomi Corporation (Author: Mingshuang Luo)
+# Copyright 2022 Xiaomi Corporation (Author: Xiaoyu Yang)
 #
 # See ../../../../LICENSE for clarification regarding multiple authors
 #
@@ -91,6 +92,22 @@ When training with the L subset, the streaming usage:
         --causal-convolution 1 \
         --decode-chunk-size 16 \
         --left-context 64
+        
+(4) modified beam search with RNNLM shallow fusion
+./pruned_transducer_stateless5/decode.py \
+    --epoch 35 \
+    --avg 15 \
+    --exp-dir ./pruned_transducer_stateless5/exp \
+    --max-duration 600 \
+    --decoding-method modified_beam_search_lm_shallow_fusion \
+    --beam-size 4 \
+    --lm-type rnn \
+    --lm-scale 0.3 \
+    --lm-exp-dir /path/to/LM \
+    --rnn-lm-epoch 99 \
+    --rnn-lm-avg 1 \
+    --rnn-lm-num-layers 3 \
+    --rnn-lm-tie-weights 1
 """
 
 
@@ -111,9 +128,12 @@ from beam_search import (
     greedy_search,
     greedy_search_batch,
     modified_beam_search,
+    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,
@@ -224,6 +244,16 @@ def get_parser():
         Used only when --decoding-method is fast_beam_search""",
     )
 
+    parser.add_argument(
+        "--ngram-lm-scale",
+        type=float,
+        default=0.01,
+        help="""
+        Used only when --decoding_method is fast_beam_search_nbest_LG and fast_beam_search_LG.
+        It specifies the scale for n-gram LM scores.
+        """,
+    )
+
     parser.add_argument(
         "--max-contexts",
         type=int,
@@ -277,6 +307,50 @@ def get_parser():
         help="left context can be seen during decoding (in frames after subsampling)",
     )
 
+    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=3,
+        help="""Token Ngram used for rescoring.
+            Used only when the decoding method is
+            modified_beam_search_ngram_rescoring, or LODR
+            """,
+    )
+
+    parser.add_argument(
+        "--backoff-id",
+        type=int,
+        default=500,
+        help="""ID of the backoff symbol.
+                Used only when the decoding method is
+                modified_beam_search_ngram_rescoring""",
+    )
     add_model_arguments(parser)
 
     return parser
@@ -288,6 +362,9 @@ def decode_one_batch(
     lexicon: Lexicon,
     batch: dict,
     decoding_graph: Optional[k2.Fsa] = None,
+    ngram_lm: Optional[NgramLm] = None,
+    ngram_lm_scale: float = 1.0,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[List[str]]]:
     """Decode one batch and return the result in a dict. The dict has the
     following format:
@@ -374,6 +451,28 @@ def decode_one_batch(
         )
         for i in range(encoder_out.size(0)):
             hyps.append([lexicon.token_table[idx] for idx in hyp_tokens[i]])
+    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 i in range(encoder_out.size(0)):
+            hyps.append([lexicon.token_table[idx] for idx in hyp_tokens[i]])
+    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 i in range(encoder_out.size(0)):
+            hyps.append([lexicon.token_table[idx] for idx in hyp_tokens[i]])
     else:
         batch_size = encoder_out.size(0)
 
@@ -419,6 +518,9 @@ def decode_dataset(
     model: nn.Module,
     lexicon: Lexicon,
     decoding_graph: Optional[k2.Fsa] = None,
+    ngram_lm: Optional[NgramLm] = None,
+    ngram_lm_scale: float = 1.0,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
     """Decode dataset.
 
@@ -432,6 +534,8 @@ def decode_dataset(
       decoding_graph:
         The decoding graph. Can be either a `k2.trivial_graph` or HLG, Used
         only when --decoding_method is fast_beam_search.
+      LM:
+        A neural network LM, used during shallow fusion
     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.
@@ -449,7 +553,7 @@ def decode_dataset(
     if params.decoding_method == "greedy_search":
         log_interval = 100
     else:
-        log_interval = 2
+        log_interval = 20
 
     results = defaultdict(list)
     for batch_idx, batch in enumerate(dl):
@@ -463,6 +567,9 @@ def decode_dataset(
             lexicon=lexicon,
             decoding_graph=decoding_graph,
             batch=batch,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+            LM=LM,
         )
 
         for name, hyps in hyps_dict.items():
@@ -524,6 +631,7 @@ def save_results(
 def main():
     parser = get_parser()
     WenetSpeechAsrDataModule.add_arguments(parser)
+    LmScorer.add_arguments(parser)
     args = parser.parse_args()
     args.exp_dir = Path(args.exp_dir)
 
@@ -535,6 +643,8 @@ def main():
         "beam_search",
         "fast_beam_search",
         "modified_beam_search",
+        "modified_beam_search_lm_shallow_fusion",
+        "modified_beam_search_LODR",
     )
     params.res_dir = params.exp_dir / params.decoding_method
 
@@ -549,6 +659,22 @@ def main():
         params.suffix += f"-context-{params.context_size}"
         params.suffix += f"-max-sym-per-frame-{params.max_sym_per_frame}"
 
+    if "ngram" in params.decoding_method:
+        params.suffix += f"-ngram-lm-scale-{params.ngram_lm_scale}"
+    if params.use_shallow_fusion:
+        if params.lm_type == "rnn":
+            params.suffix += f"-rnnlm-lm-scale-{params.lm_scale}"
+        elif params.lm_type == "transformer":
+            params.suffix += f"-transformer-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")
 
@@ -558,6 +684,7 @@ def main():
 
     logging.info(f"Device: {device}")
 
+    # import pdb; pdb.set_trace()
     lexicon = Lexicon(params.lang_dir)
     params.blank_id = lexicon.token_table["<blk>"]
     params.vocab_size = max(lexicon.tokens) + 1
@@ -652,6 +779,37 @@ def main():
     model.to(device)
     model.eval()
     model.device = device
+    # only load N-gram LM when needed
+    if "ngram" in params.decoding_method or "LODR" in params.decoding_method:
+        lm_filename = f"{params.tokens_ngram}gram.fst.txt"
+        logging.info(f"lm filename: {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
+
+    # import pdb; pdb.set_trace()
+    # only load the neural network LM if doing shallow fusion
+    if params.use_shallow_fusion:
+        LM = LmScorer(
+            lm_type=params.lm_type,
+            params=params,
+            device=device,
+            lm_scale=params.lm_scale,
+        )
+        LM.to(device)
+        LM.eval()
+
+        num_param = sum([p.numel() for p in LM.parameters()])
+        logging.info(f"Number of model parameters: {num_param}")
+    else:
+        LM = None
 
     if params.decoding_method == "fast_beam_search":
         decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)
@@ -684,6 +842,9 @@ def main():
             model=model,
             lexicon=lexicon,
             decoding_graph=decoding_graph,
+            ngram_lm=ngram_lm,
+            ngram_lm_scale=ngram_lm_scale,
+            LM=LM,
         )
         save_results(
             params=params,

icefall/lm_wrapper.py

@@ -50,7 +50,7 @@ class LmScorer(torch.nn.Module):
     def add_arguments(cls, parser):
         # LM general arguments
         parser.add_argument(
-            "--vocab-size",
+            "--lm-vocab-size",
             type=int,
             default=500,
         )

icefall/rnn_lm/compute_perplexity.py

@@ -33,7 +33,7 @@ import torch
 from dataset import get_dataloader
 from model import RnnLmModel
 
-from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.checkpoint import average_checkpoints, find_checkpoints, load_checkpoint
 from icefall.utils import AttributeDict, setup_logger, str2bool
 
 
@@ -49,6 +49,7 @@ def get_parser():
         help="It specifies the checkpoint to use for decoding."
         "Note: Epoch counts from 0.",
     )
+
     parser.add_argument(
         "--avg",
         type=int,
@@ -58,6 +59,16 @@ def get_parser():
         "'--epoch'. ",
     )
 
+    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(
         "--exp-dir",
         type=str,
@@ -154,7 +165,14 @@ def main():
 
     params = AttributeDict(vars(args))
 
-    setup_logger(f"{params.exp_dir}/log-ppl/")
+    if params.iter > 0:
+        setup_logger(
+            f"{params.exp_dir}/log-ppl/log-ppl-iter-{params.iter}-avg-{params.avg}"
+        )
+    else:
+        setup_logger(
+            f"{params.exp_dir}/log-ppl/log-ppl-epoch-{params.epoch}-avg-{params.avg}"
+        )
     logging.info("Computing perplexity started")
     logging.info(params)
 
@@ -173,19 +191,39 @@ def main():
         tie_weights=params.tie_weights,
     )
 
-    if params.avg == 1:
+    if params.iter > 0:
-        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+            : params.avg
+        ]
+        if len(filenames) == 0:
+            raise ValueError(
+                f"No checkpoints found for --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), strict=False
+        )
+    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:
+            if i >= 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))
+        model.load_state_dict(
+            average_checkpoints(filenames, device=device), strict=False
+        )
 
+    model.to(device)
     model.eval()
     num_param = sum([p.numel() for p in model.parameters()])
     num_param_requires_grad = sum(

icefall/rnn_lm/export.py

@@ -25,7 +25,7 @@ from pathlib import Path
 import torch
 from model import RnnLmModel
 
-from icefall.checkpoint import load_checkpoint
+from icefall.checkpoint import average_checkpoints, find_checkpoints, load_checkpoint
 from icefall.utils import AttributeDict, load_averaged_model, str2bool
 
 
@@ -51,6 +51,16 @@ def get_parser():
         "'--epoch'. ",
     )
 
+    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(
         "--vocab-size",
         type=int,
@@ -133,11 +143,36 @@ def main():
 
     model.to(device)
 
-    if params.avg == 1:
+    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 --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), strict=False
+        )
+    elif params.avg == 1:
         load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
     else:
-        model = load_averaged_model(
+        start = params.epoch - params.avg + 1
-            params.exp_dir, model, params.epoch, params.avg, device
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if i >= 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), strict=False
         )
 
     model.to("cpu")

icefall/rnn_lm/train.py

@@ -49,6 +49,7 @@ from torch.utils.tensorboard import SummaryWriter
 
 from icefall.checkpoint import load_checkpoint
 from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.checkpoint import save_checkpoint_with_global_batch_idx
 from icefall.dist import cleanup_dist, setup_dist
 from icefall.env import get_env_info
 from icefall.utils import AttributeDict, MetricsTracker, setup_logger, str2bool
@@ -178,6 +179,33 @@ def get_parser():
         help="The seed for random generators intended for reproducibility",
     )
 
+    parser.add_argument(
+        "--lr",
+        type=float,
+        default=1e-3,
+    )
+
+    parser.add_argument(
+        "--max-sent-len",
+        type=int,
+        default=200,
+        help="""Maximum number of tokens in a sentence. This is used
+        to adjust batch-size dynamically""",
+    )
+
+    parser.add_argument(
+        "--save-every-n",
+        type=int,
+        default=2000,
+        help="""Save checkpoint after processing this number of batches"
+        periodically. We save checkpoint to exp-dir/ whenever
+        params.batch_idx_train % save_every_n == 0. The checkpoint filename
+        has the form: f'exp-dir/checkpoint-{params.batch_idx_train}.pt'
+        Note: It also saves checkpoint to `exp-dir/epoch-xxx.pt` at the
+        end of each epoch where `xxx` is the epoch number counting from 0.
+        """,
+    )
+
     return parser
 
 
@@ -190,16 +218,15 @@ def get_params() -> AttributeDict:
             "sos_id": 1,
             "eos_id": 1,
             "blank_id": 0,
-            "lr": 1e-3,
             "weight_decay": 1e-6,
             "best_train_loss": float("inf"),
             "best_valid_loss": float("inf"),
             "best_train_epoch": -1,
             "best_valid_epoch": -1,
             "batch_idx_train": 0,
-            "log_interval": 200,
+            "log_interval": 100,
             "reset_interval": 2000,
-            "valid_interval": 5000,
+            "valid_interval": 200,
             "env_info": get_env_info(),
         }
     )
@@ -382,6 +409,7 @@ def train_one_epoch(
     valid_dl: torch.utils.data.DataLoader,
     tb_writer: Optional[SummaryWriter] = None,
     world_size: int = 1,
+    rank: int = 0,
 ) -> None:
     """Train the model for one epoch.
 
@@ -430,6 +458,19 @@ def train_one_epoch(
         clip_grad_norm_(model.parameters(), 5.0, 2.0)
         optimizer.step()
 
+        if (
+            params.batch_idx_train > 0
+            and params.batch_idx_train % params.save_every_n == 0
+        ):
+            save_checkpoint_with_global_batch_idx(
+                out_dir=params.exp_dir,
+                global_batch_idx=params.batch_idx_train,
+                model=model,
+                params=params,
+                optimizer=optimizer,
+                rank=rank,
+            )
+
         if batch_idx % params.log_interval == 0:
             # Note: "frames" here means "num_tokens"
             this_batch_ppl = math.exp(loss_info["loss"] / loss_info["frames"])
@@ -580,6 +621,7 @@ def run(rank, world_size, args):
             valid_dl=valid_dl,
             tb_writer=tb_writer,
             world_size=world_size,
+            rank=rank,
         )
 
         save_checkpoint(

icefall/utils.py

@@ -1378,7 +1378,7 @@ def parse_timestamp(tokens: List[str], timestamp: List[float]) -> List[float]:
       List of timestamp of each word.
     """
     start_token = b"\xe2\x96\x81".decode()  # '_'
-    assert len(tokens) == len(timestamp)
+    assert len(tokens) == len(timestamp), (len(tokens), len(timestamp))
     ans = []
     for i in range(len(tokens)):
         flag = False