SPGISpeech recipe (#334)

* initial commit for SPGISpeech recipe

* add decoding

* add spgispeech transducer

* remove conformer ctc; minor fixes in RNN-T

* add results

* add tensorboard

* add pretrained model to HF

* remove unused scripts and soft link common scripts

* remove duplicate files

* pre commit hooks

* remove change in librispeech

* pre commit hook

* add CER numbers

egs/spgispeech/ASR/README.md

@@ -0,0 +1,32 @@
+# SPGISpeech
+
+SPGISpeech consists of 5,000 hours of recorded company earnings calls and their respective
+transcriptions. The original calls were split into slices ranging from 5 to 15 seconds in
+length to allow easy training for speech recognition systems. Calls represent a broad
+cross-section of international business English; SPGISpeech contains approximately 50,000
+speakers, one of the largest numbers of any speech corpus, and offers a variety of L1 and
+L2 English accents. The format of each WAV file is single channel, 16kHz, 16 bit audio.
+
+Transcription text represents the output of several stages of manual post-processing.
+As such, the text contains polished English orthography following a detailed style guide,
+including proper casing, punctuation, and denormalized non-standard words such as numbers
+and acronyms, making SPGISpeech suited for training fully formatted end-to-end models.
+
+Official reference:
+
+O’Neill, P.K., Lavrukhin, V., Majumdar, S., Noroozi, V., Zhang, Y., Kuchaiev, O., Balam,
+J., Dovzhenko, Y., Freyberg, K., Shulman, M.D., Ginsburg, B., Watanabe, S., & Kucsko, G.
+(2021). SPGISpeech: 5, 000 hours of transcribed financial audio for fully formatted
+end-to-end speech recognition. ArXiv, abs/2104.02014.
+
+ArXiv link: https://arxiv.org/abs/2104.02014
+
+## Performance Record
+
+| Decoding method           | val WER    | val CER |
+|---------------------------|------------|---------|
+| greedy search             | 2.40       |  0.99   |
+| modified beam search      | 2.24       |  0.91   |
+| fast beam search          | 2.35       |  0.97   |
+
+See [RESULTS](/egs/spgispeech/ASR/RESULTS.md) for details.

egs/spgispeech/ASR/RESULTS.md

@@ -0,0 +1,73 @@
+## Results
+
+### SPGISpeech BPE training results (Pruned Transducer)
+
+#### 2022-05-11
+
+#### Conformer encoder + embedding decoder
+
+Conformer encoder + non-current decoder. The decoder
+contains only an embedding layer, a Conv1d (with kernel size 2) and a linear
+layer (to transform tensor dim).
+
+The WERs are
+
+|                           | dev | val | comment                                  |
+|---------------------------|------------|------------|------------------------------------------|
+| greedy search             | 2.46       | 2.40       | --avg-last-n 10 --max-duration 500 |
+| modified beam search      | 2.28       | 2.24       | --avg-last-n 10 --max-duration 500 --beam-size 4 |
+| fast beam search          | 2.38       | 2.35       | --avg-last-n 10 --max-duration 500 --beam-size 4 --max-contexts 4 --max-states 8 |
+
+**NOTE:** SPGISpeech transcripts can be prepared in `ortho` or `norm` ways, which refer to whether the
+transcripts are orthographic or normalized. These WERs correspond to the normalized transcription
+scenario.
+
+The training command for reproducing is given below:
+
+```
+export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"
+
+./pruned_transducer_stateless2/train.py \
+  --world-size 8 \
+  --num-epochs 20 \
+  --start-epoch 0 \
+  --exp-dir pruned_transducer_stateless2/exp \
+  --max-duration 200 \
+  --prune-range 5 \
+  --lr-factor 5 \
+  --lm-scale 0.25 \
+  --use-fp16 True
+```
+
+The decoding command is:
+```
+# greedy search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 100 \
+        --decoding-method greedy_search
+
+# modified beam search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 100 \
+        --decoding-method modified_beam_search \
+        --beam-size 4
+
+# fast beam search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 1500 \
+        --decoding-method fast_beam_search \
+        --beam 4 \
+        --max-contexts 4 \
+        --max-states 8
+```
+
+Pretrained model is available at <https://huggingface.co/desh2608/icefall-asr-spgispeech-pruned-transducer-stateless2>
+
+The tensorboard training log can be found at
+<https://tensorboard.dev/experiment/ExSoBmrPRx6liMTGLu0Tgw/#scalars>

egs/spgispeech/ASR/local/compile_hlg.py

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

egs/spgispeech/ASR/local/compute_fbank_musan.py

@@ -0,0 +1,104 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This file computes fbank features of the musan dataset.
+It looks for manifests in the directory data/manifests.
+
+The generated fbank features are saved in data/fbank.
+"""
+
+import logging
+from pathlib import Path
+
+import torch
+from lhotse import LilcomChunkyWriter, CutSet, combine
+from lhotse.features.kaldifeat import (
+    KaldifeatFbank,
+    KaldifeatFbankConfig,
+    KaldifeatMelOptions,
+    KaldifeatFrameOptions,
+)
+from lhotse.recipes.utils import read_manifests_if_cached
+
+from icefall.utils import get_executor
+
+# 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 compute_fbank_musan():
+    src_dir = Path("data/manifests")
+    output_dir = Path("data/fbank")
+
+    sampling_rate = 16000
+    num_mel_bins = 80
+
+    extractor = KaldifeatFbank(
+        KaldifeatFbankConfig(
+            frame_opts=KaldifeatFrameOptions(sampling_rate=sampling_rate),
+            mel_opts=KaldifeatMelOptions(num_bins=num_mel_bins),
+            device="cuda",
+        )
+    )
+
+    dataset_parts = (
+        "music",
+        "speech",
+        "noise",
+    )
+    manifests = read_manifests_if_cached(
+        dataset_parts=dataset_parts, output_dir=src_dir
+    )
+    assert manifests is not None
+
+    musan_cuts_path = src_dir / "cuts_musan.jsonl.gz"
+
+    if musan_cuts_path.is_file():
+        logging.info(f"{musan_cuts_path} already exists - skipping")
+        return
+
+    logging.info("Extracting features for Musan")
+
+    # create chunks of Musan with duration 5 - 10 seconds
+    musan_cuts = (
+        CutSet.from_manifests(
+            recordings=combine(part["recordings"] for part in manifests.values())
+        )
+        .cut_into_windows(10.0)
+        .filter(lambda c: c.duration > 5)
+        .compute_and_store_features_batch(
+            extractor=extractor,
+            storage_path=output_dir / f"feats_musan",
+            manifest_path=src_dir / f"cuts_musan.jsonl.gz",
+            batch_duration=500,
+            num_workers=4,
+            storage_type=LilcomChunkyWriter,
+        )
+    )
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    compute_fbank_musan()

egs/spgispeech/ASR/local/compute_fbank_spgispeech.py

@@ -0,0 +1,145 @@
+#!/usr/bin/env python3
+# Copyright    2022  Johns Hopkins University        (authors: Desh Raj)
+#
+# 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 SPGISpeech dataset.
+It looks for manifests in the directory data/manifests.
+
+The generated fbank features are saved in data/fbank.
+"""
+import argparse
+import logging
+from pathlib import Path
+from tqdm import tqdm
+
+import torch
+from lhotse import load_manifest_lazy, LilcomChunkyWriter
+from lhotse.features.kaldifeat import (
+    KaldifeatFbank,
+    KaldifeatFbankConfig,
+    KaldifeatMelOptions,
+    KaldifeatFrameOptions,
+)
+from lhotse.manipulation import combine
+
+# 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(
+        "--num-splits",
+        type=int,
+        default=20,
+        help="Number of splits for the train set.",
+    )
+    parser.add_argument(
+        "--start",
+        type=int,
+        default=0,
+        help="Start index of the train set split.",
+    )
+    parser.add_argument(
+        "--stop",
+        type=int,
+        default=-1,
+        help="Stop index of the train set split.",
+    )
+    parser.add_argument(
+        "--test",
+        action="store_true",
+        help="If set, only compute features for the dev and val set.",
+    )
+    parser.add_argument(
+        "--train",
+        action="store_true",
+        help="If set, only compute features for the train set.",
+    )
+
+    return parser.parse_args()
+
+
+def compute_fbank_spgispeech(args):
+    assert args.train or args.test, "Either train or test must be set."
+
+    src_dir = Path("data/manifests")
+    output_dir = Path("data/fbank")
+
+    sampling_rate = 16000
+    num_mel_bins = 80
+
+    extractor = KaldifeatFbank(
+        KaldifeatFbankConfig(
+            frame_opts=KaldifeatFrameOptions(sampling_rate=sampling_rate),
+            mel_opts=KaldifeatMelOptions(num_bins=num_mel_bins),
+            device="cuda",
+        )
+    )
+
+    if args.train:
+        logging.info(f"Processing train")
+        cut_set = load_manifest_lazy(src_dir / f"cuts_train_raw.jsonl.gz")
+        chunk_size = len(cut_set) // args.num_splits
+        cut_sets = cut_set.split_lazy(
+            output_dir=src_dir / f"cuts_train_raw_split{args.num_splits}",
+            chunk_size=chunk_size,
+        )
+        start = args.start
+        stop = min(args.stop, args.num_splits) if args.stop > 0 else args.num_splits
+        num_digits = len(str(args.num_splits))
+        for i in range(start, stop):
+            idx = f"{i + 1}".zfill(num_digits)
+            logging.info(f"Processing train split {i}")
+            cs = cut_sets[i].compute_and_store_features_batch(
+                extractor=extractor,
+                storage_path=output_dir / f"feats_train_{idx}",
+                manifest_path=src_dir / f"cuts_train_{idx}.jsonl.gz",
+                batch_duration=500,
+                num_workers=4,
+                storage_type=LilcomChunkyWriter,
+            )
+
+    if args.test:
+        for partition in ["dev", "val"]:
+            if (output_dir / f"cuts_{partition}.jsonl.gz").is_file():
+                logging.info(f"{partition} already exists - skipping.")
+                continue
+            logging.info(f"Processing {partition}")
+            cut_set = load_manifest_lazy(src_dir / f"cuts_{partition}_raw.jsonl.gz")
+            cut_set = cut_set.compute_and_store_features_batch(
+                extractor=extractor,
+                storage_path=output_dir / f"feats_{partition}",
+                manifest_path=src_dir / f"cuts_{partition}.jsonl.gz",
+                batch_duration=500,
+                num_workers=4,
+                storage_type=LilcomChunkyWriter,
+            )
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    args = get_args()
+    compute_fbank_spgispeech(args)

egs/spgispeech/ASR/local/prepare_lang.py

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

egs/spgispeech/ASR/local/prepare_lang_bpe.py

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

egs/spgispeech/ASR/local/prepare_splits.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python3
+# Copyright    2022  Johns Hopkins University        (authors: Desh Raj)
+#
+# 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 splits the training set into train and dev sets.
+"""
+import logging
+from pathlib import Path
+
+import torch
+from lhotse import CutSet
+
+from lhotse.recipes.utils import read_manifests_if_cached
+
+# 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 split_spgispeech_train():
+    src_dir = Path("data/manifests")
+
+    manifests = read_manifests_if_cached(
+        dataset_parts=["train", "val"],
+        output_dir=src_dir,
+        prefix="spgispeech",
+        suffix="jsonl.gz",
+        lazy=True,
+    )
+    assert manifests is not None
+
+    train_dev_cuts = CutSet.from_manifests(
+        recordings=manifests["train"]["recordings"],
+        supervisions=manifests["train"]["supervisions"],
+    )
+    dev_cuts = train_dev_cuts.subset(first=4000)
+    train_cuts = train_dev_cuts.filter(lambda c: c not in dev_cuts)
+
+    # Add speed perturbation
+    train_cuts = (
+        train_cuts + train_cuts.perturb_speed(0.9) + train_cuts.perturb_speed(1.1)
+    )
+
+    # Write the manifests to disk.
+    train_cuts.to_file(src_dir / "cuts_train_raw.jsonl.gz")
+    dev_cuts.to_file(src_dir / "cuts_dev_raw.jsonl.gz")
+
+    # Also write the val set to disk.
+    val_cuts = CutSet.from_manifests(
+        recordings=manifests["val"]["recordings"],
+        supervisions=manifests["val"]["supervisions"],
+    )
+    val_cuts.to_file(src_dir / "cuts_val_raw.jsonl.gz")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+
+    split_spgispeech_train()

egs/spgispeech/ASR/local/train_bpe_model.py

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

egs/spgispeech/ASR/prepare.sh

@@ -0,0 +1,196 @@
+#!/usr/bin/env bash
+
+set -eou pipefail
+
+nj=20
+stage=-1
+stop_stage=100
+
+# We assume dl_dir (download dir) contains the following
+# directories and files. If not, they will be downloaded
+# by this script automatically.
+#
+#  - $dl_dir/spgispeech
+#      You can find train.csv, val.csv, train, and val in this directory, which belong
+#      to the SPGISpeech dataset.
+#
+#  - $dl_dir/musan
+#      This directory contains the following directories downloaded from
+#       http://www.openslr.org/17/
+#
+#     - music
+#     - noise
+#     - speech
+dl_dir=$PWD/download
+
+. shared/parse_options.sh || exit 1
+
+# vocab size for sentence piece models.
+# It will generate data/lang_bpe_xxx,
+# data/lang_bpe_yyy if the array contains xxx, yyy
+vocab_sizes=(
+  500
+)
+
+# All files generated by this script are saved in "data".
+# You can safely remove "data" and rerun this script to regenerate it.
+mkdir -p data
+
+log() {
+  # This function is from espnet
+  local fname=${BASH_SOURCE[1]##*/}
+  echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+log "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/spgispeech,
+  # you can create a symlink
+  #
+  #   ln -sfv /path/to/spgispeech $dl_dir/spgispeech
+  #
+  if [ ! -d $dl_dir/spgispeech/train.csv ]; then
+    lhotse download spgispeech $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 SPGISpeech manifest (may take ~1h)"
+  # We assume that you have downloaded the SPGISpeech corpus
+  # to $dl_dir/spgispeech. We perform text normalization for the transcripts.
+  mkdir -p data/manifests
+  lhotse prepare spgispeech -j $nj --normalize-text $dl_dir/spgispeech data/manifests
+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
+  lhotse prepare musan $dl_dir/musan data/manifests
+  lhotse combine data/manifests/recordings_{music,speech,noise}.json data/manifests/recordings_musan.jsonl.gz
+  lhotse cut simple -r data/manifests/recordings_musan.jsonl.gz data/manifests/cuts_musan_raw.jsonl.gz
+fi
+
+if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
+  log "Stage 3: Split train into train and dev and create cut sets."
+  python local/prepare_splits.py
+fi
+
+if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
+  log "Stage 4: Compute fbank features for spgispeech dev and val"
+  mkdir -p data/fbank
+  python local/compute_fbank_spgispeech.py --test
+fi
+
+if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
+  log "Stage 5: Compute fbank features for train"
+  mkdir -p data/fbank
+  python local/compute_fbank_spgispeech.py --train --num-splits 20
+
+  log "Combine features from train splits (may take ~1h)"
+  if [ ! -f data/manifests/cuts_train.jsonl.gz ]; then
+    pieces=$(find data/manifests -name "cuts_train_[0-9]*.jsonl.gz")
+    lhotse combine $pieces data/manifests/cuts_train.jsonl.gz
+  fi
+  gunzip -c data/manifests/train_cuts.jsonl.gz | shuf | gzip -c > data/manifests/train_cuts_shuf.jsonl.gz
+fi
+
+if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
+  log "Stage 6: Compute fbank features for musan"
+  mkdir -p data/fbank
+  python local/compute_fbank_musan.py
+fi
+
+if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
+  log "Stage 7: Dump transcripts for LM training"
+  mkdir -p data/lm
+  gunzip -c data/manifests/cuts_train_raw.jsonl.gz \
+    | jq '.supervisions[0].text' \
+    | sed 's:"::g' \
+    > data/lm/transcript_words.txt
+fi
+
+if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
+  log "Stage 8: Prepare BPE based lang"
+
+  for vocab_size in ${vocab_sizes[@]}; do
+    lang_dir=data/lang_bpe_${vocab_size}
+    mkdir -p $lang_dir
+
+    # Add special words to words.txt
+    echo "<eps> 0" > $lang_dir/words.txt
+    echo "!SIL 1" >> $lang_dir/words.txt
+    echo "[UNK] 2" >> $lang_dir/words.txt
+
+    # Add regular words to words.txt
+    gunzip -c data/manifests/cuts_train_raw.jsonl.gz \
+      | jq '.supervisions[0].text' \
+      | sed 's:"::g' \
+      | sed 's: :\n:g' \
+      | sort \
+      | uniq \
+      | sed '/^$/d' \
+      | awk '{print $0,NR+2}' \
+      >> $lang_dir/words.txt
+
+    # Add remaining special word symbols expected by LM scripts.
+    num_words=$(cat $lang_dir/words.txt | wc -l)
+    echo "<s> ${num_words}" >> $lang_dir/words.txt
+    num_words=$(cat $lang_dir/words.txt | wc -l)
+    echo "</s> ${num_words}" >> $lang_dir/words.txt
+    num_words=$(cat $lang_dir/words.txt | wc -l)
+    echo "#0 ${num_words}" >> $lang_dir/words.txt
+
+    ./local/train_bpe_model.py \
+      --lang-dir $lang_dir \
+      --vocab-size $vocab_size \
+      --transcript data/lm/transcript_words.txt
+
+    if [ ! -f $lang_dir/L_disambig.pt ]; then
+      ./local/prepare_lang_bpe.py --lang-dir $lang_dir
+    fi
+  done
+fi
+
+if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
+  log "Stage 9: Train LM"
+  lm_dir=data/lm
+
+  if [ ! -f $lm_dir/G.arpa ]; then
+    ./shared/make_kn_lm.py \
+      -ngram-order 3 \
+      -text $lm_dir/transcript_words.txt \
+      -lm $lm_dir/G.arpa
+  fi
+
+  if [ ! -f $lm_dir/G_3_gram.fst.txt ]; then
+    python3 -m kaldilm \
+      --read-symbol-table="data/lang_phone/words.txt" \
+      --disambig-symbol='#0' \
+      --max-order=3 \
+      $lm_dir/G.arpa > $lm_dir/G_3_gram.fst.txt
+  fi
+fi
+
+if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
+  log "Stage 10: Compile HLG"
+  ./local/compile_hlg.py --lang-dir data/lang_phone
+
+  for vocab_size in ${vocab_sizes[@]}; do
+    lang_dir=data/lang_bpe_${vocab_size}
+    ./local/compile_hlg.py --lang-dir $lang_dir
+  done
+fi

egs/spgispeech/ASR/pruned_transducer_stateless2/asr_datamodule.py

@@ -0,0 +1,366 @@
+# Copyright      2021  Piotr Żelasko
+#
+# 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 functools import lru_cache
+from pathlib import Path
+from typing import Any, Dict, Optional
+
+import torch
+from lhotse import CutSet, Fbank, FbankConfig, load_manifest, load_manifest_lazy
+from lhotse.dataset import (
+    CutConcatenate,
+    CutMix,
+    DynamicBucketingSampler,
+    K2SpeechRecognitionDataset,
+    PrecomputedFeatures,
+    SpecAugment,
+)
+from lhotse.dataset.input_strategies import OnTheFlyFeatures
+from lhotse.utils import fix_random_seed
+from torch.utils.data import DataLoader
+from tqdm import tqdm
+
+from icefall.utils import str2bool
+
+
+class _SeedWorkers:
+    def __init__(self, seed: int):
+        self.seed = seed
+
+    def __call__(self, worker_id: int):
+        fix_random_seed(self.seed + worker_id)
+
+
+class SPGISpeechAsrDataModule:
+    """
+    DataModule for k2 ASR experiments.
+    It assumes there is always one train and valid dataloader,
+    but there can be multiple test dataloaders (e.g. LibriSpeech test-clean
+    and test-other).
+    It contains all the common data pipeline modules used in ASR
+    experiments, e.g.:
+    - dynamic batch size,
+    - bucketing samplers,
+    - cut concatenation,
+    - augmentation,
+    - on-the-fly feature extraction
+    This class should be derived for specific corpora used in ASR tasks.
+    """
+
+    def __init__(self, args: argparse.Namespace):
+        self.args = args
+
+    @classmethod
+    def add_arguments(cls, parser: argparse.ArgumentParser):
+        group = parser.add_argument_group(
+            title="ASR data related options",
+            description="These options are used for the preparation of "
+            "PyTorch DataLoaders from Lhotse CutSet's -- they control the "
+            "effective batch sizes, sampling strategies, applied data "
+            "augmentations, etc.",
+        )
+        group.add_argument(
+            "--manifest-dir",
+            type=Path,
+            default=Path("data/manifests"),
+            help="Path to directory with train/valid/test cuts.",
+        )
+        group.add_argument(
+            "--enable-musan",
+            type=str2bool,
+            default=True,
+            help="When enabled, select noise from MUSAN and mix it "
+            "with training dataset. ",
+        )
+        group.add_argument(
+            "--concatenate-cuts",
+            type=str2bool,
+            default=False,
+            help="When enabled, utterances (cuts) will be concatenated "
+            "to minimize the amount of padding.",
+        )
+        group.add_argument(
+            "--duration-factor",
+            type=float,
+            default=1.0,
+            help="Determines the maximum duration of a concatenated cut "
+            "relative to the duration of the longest cut in a batch.",
+        )
+        group.add_argument(
+            "--gap",
+            type=float,
+            default=1.0,
+            help="The amount of padding (in seconds) inserted between "
+            "concatenated cuts. This padding is filled with noise when "
+            "noise augmentation is used.",
+        )
+        group.add_argument(
+            "--max-duration",
+            type=int,
+            default=100.0,
+            help="Maximum pooled recordings duration (seconds) in a "
+            "single batch. You can reduce it if it causes CUDA OOM.",
+        )
+        group.add_argument(
+            "--num-buckets",
+            type=int,
+            default=30,
+            help="The number of buckets for the BucketingSampler"
+            "(you might want to increase it for larger datasets).",
+        )
+        group.add_argument(
+            "--on-the-fly-feats",
+            type=str2bool,
+            default=False,
+            help="When enabled, use on-the-fly cut mixing and feature "
+            "extraction. Will drop existing precomputed feature manifests "
+            "if available.",
+        )
+        group.add_argument(
+            "--shuffle",
+            type=str2bool,
+            default=True,
+            help="When enabled (=default), the examples will be "
+            "shuffled for each epoch.",
+        )
+
+        group.add_argument(
+            "--num-workers",
+            type=int,
+            default=8,
+            help="The number of training dataloader workers that "
+            "collect the batches.",
+        )
+        group.add_argument(
+            "--enable-spec-aug",
+            type=str2bool,
+            default=True,
+            help="When enabled, use SpecAugment for training dataset.",
+        )
+        group.add_argument(
+            "--spec-aug-time-warp-factor",
+            type=int,
+            default=80,
+            help="Used only when --enable-spec-aug is True. "
+            "It specifies the factor for time warping in SpecAugment. "
+            "Larger values mean more warping. "
+            "A value less than 1 means to disable time warp.",
+        )
+
+    def train_dataloaders(
+        self,
+        cuts_train: CutSet,
+        sampler_state_dict: Optional[Dict[str, Any]] = None,
+    ) -> DataLoader:
+        """
+        Args:
+          cuts_train:
+            CutSet for training.
+          sampler_state_dict:
+            The state dict for the training sampler.
+        """
+        logging.info("About to get Musan cuts")
+        cuts_musan = load_manifest(
+            self.args.manifest_dir / "cuts_musan.jsonl.gz"
+        )
+
+        transforms = []
+        if self.args.enable_musan:
+            logging.info("Enable MUSAN")
+            transforms.append(
+                CutMix(
+                    cuts=cuts_musan, prob=0.5, snr=(10, 20), preserve_id=True
+                )
+            )
+        else:
+            logging.info("Disable MUSAN")
+
+        if self.args.concatenate_cuts:
+            logging.info(
+                f"Using cut concatenation with duration factor "
+                f"{self.args.duration_factor} and gap {self.args.gap}."
+            )
+            # Cut concatenation should be the first transform in the list,
+            # so that if we e.g. mix noise in, it will fill the gaps between
+            # different utterances.
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        input_transforms = []
+        if self.args.enable_spec_aug:
+            logging.info("Enable SpecAugment")
+            logging.info(
+                f"Time warp factor: {self.args.spec_aug_time_warp_factor}"
+            )
+            input_transforms.append(
+                SpecAugment(
+                    time_warp_factor=self.args.spec_aug_time_warp_factor,
+                    num_frame_masks=2,
+                    features_mask_size=27,
+                    num_feature_masks=2,
+                    frames_mask_size=100,
+                )
+            )
+        else:
+            logging.info("Disable SpecAugment")
+
+        logging.info("About to create train dataset")
+        if self.args.on_the_fly_feats:
+            train = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(
+                    Fbank(FbankConfig(num_mel_bins=80))
+                ),
+                input_transforms=input_transforms,
+            )
+        else:
+            train = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_transforms=input_transforms,
+            )
+
+        logging.info("Using DynamicBucketingSampler.")
+        train_sampler = DynamicBucketingSampler(
+            cuts_train,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+            num_buckets=self.args.num_buckets,
+            drop_last=True,
+        )
+        logging.info("About to create train dataloader")
+
+        if sampler_state_dict is not None:
+            logging.info("Loading sampler state dict")
+            train_sampler.load_state_dict(sampler_state_dict)
+
+        # 'seed' is derived from the current random state, which will have
+        # previously been set in the main process.
+        seed = torch.randint(0, 100000, ()).item()
+        worker_init_fn = _SeedWorkers(seed)
+
+        train_dl = DataLoader(
+            train,
+            sampler=train_sampler,
+            batch_size=None,
+            num_workers=self.args.num_workers,
+            persistent_workers=False,
+            worker_init_fn=worker_init_fn,
+        )
+
+        return train_dl
+
+    def valid_dataloaders(self, cuts_valid: CutSet) -> DataLoader:
+
+        transforms = []
+        if self.args.concatenate_cuts:
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        logging.info("About to create dev dataset")
+        if self.args.on_the_fly_feats:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(
+                    Fbank(FbankConfig(num_mel_bins=80))
+                ),
+            )
+        else:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+            )
+        valid_sampler = DynamicBucketingSampler(
+            cuts_valid,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+        )
+        logging.info("About to create dev dataloader")
+        valid_dl = DataLoader(
+            validate,
+            sampler=valid_sampler,
+            batch_size=None,
+            num_workers=2,
+            persistent_workers=False,
+        )
+
+        return valid_dl
+
+    def test_dataloaders(self, cuts: CutSet) -> DataLoader:
+        logging.debug("About to create test dataset")
+        test = K2SpeechRecognitionDataset(
+            input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80)))
+            if self.args.on_the_fly_feats
+            else PrecomputedFeatures(),
+        )
+        sampler = DynamicBucketingSampler(
+            cuts, max_duration=self.args.max_duration, shuffle=False
+        )
+        logging.debug("About to create test dataloader")
+        test_dl = DataLoader(
+            test,
+            batch_size=None,
+            sampler=sampler,
+            num_workers=self.args.num_workers,
+        )
+        return test_dl
+
+    @lru_cache()
+    def train_cuts(self) -> CutSet:
+        logging.info("About to get SPGISpeech train cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "cuts_train_shuf.jsonl.gz"
+        )
+
+    @lru_cache()
+    def dev_cuts(self) -> CutSet:
+        logging.info("About to get SPGISpeech dev cuts")
+        return load_manifest_lazy(self.args.manifest_dir / "cuts_dev.jsonl.gz")
+
+    @lru_cache()
+    def val_cuts(self) -> CutSet:
+        logging.info("About to get SPGISpeech val cuts")
+        return load_manifest_lazy(self.args.manifest_dir / "cuts_val.jsonl.gz")
+
+
+def test():
+    parser = argparse.ArgumentParser()
+    SPGISpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    adm = SPGISpeechAsrDataModule(args)
+
+    cuts = adm.train_cuts()
+    dl = adm.train_dataloaders(cuts)
+    for i, batch in tqdm(enumerate(dl)):
+        if i == 100:
+            break
+
+    cuts = adm.dev_cuts()
+    dl = adm.valid_dataloaders(cuts)
+    for i, batch in tqdm(enumerate(dl)):
+        if i == 100:
+            break
+
+
+if __name__ == "__main__":
+    test()

egs/spgispeech/ASR/pruned_transducer_stateless2/beam_search.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/beam_search.py

egs/spgispeech/ASR/pruned_transducer_stateless2/conformer.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/conformer.py

egs/spgispeech/ASR/pruned_transducer_stateless2/decode.py

@@ -0,0 +1,594 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+(1) greedy search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 \
+        --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 100 \
+        --decoding-method greedy_search
+
+(2) beam search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 \
+        --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 100 \
+        --decoding-method beam_search \
+        --beam-size 4
+
+(3) modified beam search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 \
+        --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 100 \
+        --decoding-method modified_beam_search \
+        --beam-size 4
+
+(4) fast beam search
+./pruned_transducer_stateless2/decode.py \
+        --iter 696000 \
+        --avg 10 \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --max-duration 1500 \
+        --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 sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import SPGISpeechAsrDataModule
+from beam_search import (
+    beam_search,
+    fast_beam_search_one_best,
+    greedy_search,
+    greedy_search_batch,
+    modified_beam_search,
+)
+from train import get_params, get_transducer_model
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    find_checkpoints,
+    load_checkpoint,
+)
+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=20,
+        help="""It specifies the checkpoint to use for decoding.
+        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=10,
+        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(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+          - modified_beam_search
+          - fast_beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=4,
+        help="""An interger 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=2,
+        help="The context size in the decoder. 1 means bigram; "
+        "2 means tri-gram",
+    )
+    parser.add_argument(
+        "--max-sym-per-frame",
+        type=int,
+        default=1,
+        help="""Maximum number of symbols per frame.
+        Used only when --decoding_method is greedy_search""",
+    )
+
+    return parser
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    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.
+      sp:
+        The BPE model.
+      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 = model.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
+    )
+    hyps = []
+
+    if params.decoding_method == "fast_beam_search":
+        hyp_tokens = fast_beam_search_one_best(
+            model=model,
+            decoding_graph=decoding_graph,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam,
+            max_contexts=params.max_contexts,
+            max_states=params.max_states,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif (
+        params.decoding_method == "greedy_search"
+        and params.max_sym_per_frame == 1
+    ):
+        hyp_tokens = greedy_search_batch(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search":
+        hyp_tokens = modified_beam_search(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    else:
+        batch_size = encoder_out.size(0)
+
+        for i in range(batch_size):
+            # fmt: off
+            encoder_out_i = encoder_out[i:i+1, :encoder_out_lens[i]]
+            # fmt: on
+            if params.decoding_method == "greedy_search":
+                hyp = greedy_search(
+                    model=model,
+                    encoder_out=encoder_out_i,
+                    max_sym_per_frame=params.max_sym_per_frame,
+                )
+            elif params.decoding_method == "beam_search":
+                hyp = beam_search(
+                    model=model,
+                    encoder_out=encoder_out_i,
+                    beam=params.beam_size,
+                )
+            else:
+                raise ValueError(
+                    f"Unsupported decoding method: {params.decoding_method}"
+                )
+            hyps.append(sp.decode(hyp).split())
+
+    if params.decoding_method == "greedy_search":
+        return {"greedy_search": hyps}
+    elif 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,
+    sp: spm.SentencePieceProcessor,
+    decoding_graph: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      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 = 100
+    else:
+        log_interval = 2
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            sp=sp,
+            decoding_graph=decoding_graph,
+            batch=batch,
+        )
+
+        for name, hyps in hyps_dict.items():
+            this_batch = []
+            assert len(hyps) == len(texts)
+            for hyp_words, ref_text in zip(hyps, texts):
+                ref_words = ref_text.split()
+                this_batch.append((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[List[int], List[int]]]],
+):
+    test_set_wers = dict()
+    test_set_cers = dict()
+    for key, results in results_dict.items():
+        recog_path = (
+            params.res_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        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.
+        wers_filename = (
+            params.res_dir / f"wers-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        with open(wers_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results, enable_log=True
+            )
+            test_set_wers[key] = wer
+
+        # we also compute CER for spgispeech dataset.
+        results_char = []
+        for res in results:
+            results_char.append((list("".join(res[0])), list("".join(res[1]))))
+        cers_filename = (
+            params.res_dir / f"cers-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        with open(cers_filename, "w") as f:
+            cer = write_error_stats(
+                f, f"{test_set_name}-{key}", results_char, enable_log=True
+            )
+            test_set_cers[key] = cer
+
+        logging.info("Wrote detailed error stats to {}".format(wers_filename))
+
+    test_set_wers = {
+        k: v for k, v in sorted(test_set_wers.items(), key=lambda x: x[1])
+    }
+    test_set_cers = {
+        k: v for k, v in sorted(test_set_cers.items(), key=lambda x: x[1])
+    }
+    errs_info = (
+        params.res_dir
+        / f"wer-summary-{test_set_name}-{key}-{params.suffix}.txt"
+    )
+    with open(errs_info, "w") as f:
+        print("settings\tWER\tCER", file=f)
+        for key in test_set_wers:
+            print(
+                "{}\t{}\t{}".format(
+                    key, test_set_wers[key], test_set_cers[key]
+                ),
+                file=f,
+            )
+
+    s = "\nFor {}, WER/CER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key in test_set_wers:
+        s += "{}\t{}\t{}{}\n".format(
+            key, test_set_wers[key], test_set_cers[key], note
+        )
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    SPGISpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    assert params.decoding_method in (
+        "greedy_search",
+        "beam_search",
+        "fast_beam_search",
+        "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}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> is 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)
+
+    if params.iter > 0:
+        filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+            : params.avg
+        ]
+        if len(filenames) == 0:
+            raise ValueError(
+                f"No checkpoints found for"
+                f" --iter {params.iter}, --avg {params.avg}"
+            )
+        elif len(filenames) < params.avg:
+            raise ValueError(
+                f"Not enough checkpoints ({len(filenames)}) found for"
+                f" --iter {params.iter}, --avg {params.avg}"
+            )
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+    elif params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+
+    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)
+    else:
+        decoding_graph = None
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    spgispeech = SPGISpeechAsrDataModule(args)
+
+    dev_cuts = spgispeech.dev_cuts()
+    val_cuts = spgispeech.val_cuts()
+
+    dev_dl = spgispeech.test_dataloaders(dev_cuts)
+    val_dl = spgispeech.test_dataloaders(val_cuts)
+
+    test_sets = ["dev", "val"]
+    test_dl = [dev_dl, val_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            sp=sp,
+            decoding_graph=decoding_graph,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    main()

egs/spgispeech/ASR/pruned_transducer_stateless2/decoder.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/decoder.py

egs/spgispeech/ASR/pruned_transducer_stateless2/encoder_interface.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/encoder_interface.py

egs/spgispeech/ASR/pruned_transducer_stateless2/export.py

@@ -0,0 +1,201 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This script converts several saved checkpoints
+# to a single one using model averaging.
+"""
+Usage:
+./pruned_transducer_stateless2/export.py \
+  --exp-dir ./pruned_transducer_stateless2/exp \
+  --bpe-model data/lang_bpe_500/bpe.model \
+  --avg-last-n 10
+
+It will generate a file exp_dir/pretrained.pt
+
+To use the generated file with `pruned_transducer_stateless2/decode.py`,
+you can do:
+
+    cd /path/to/exp_dir
+    ln -s pretrained.pt epoch-9999.pt
+
+    cd /path/to/egs/spgispeech/ASR
+    ./pruned_transducer_stateless2/decode.py \
+        --exp-dir ./pruned_transducer_stateless2/exp \
+        --epoch 9999 \
+        --avg 1 \
+        --max-duration 100 \
+        --bpe-model data/lang_bpe_500/bpe.model
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import sentencepiece as spm
+import torch
+from train import get_params, get_transducer_model
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.utils import 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 decoding."
+        "Note: Epoch counts from 0.",
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--avg-last-n",
+        type=int,
+        default=0,
+        help="""If positive, --epoch and --avg are ignored and it
+        will use the last n checkpoints exp_dir/checkpoint-xxx.pt
+        where xxx is the number of processed batches while
+        saving that checkpoint.
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="pruned_transducer_stateless2/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--jit",
+        type=str2bool,
+        default=False,
+        help="""True to save a model after applying torch.jit.script.
+        """,
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; "
+        "2 means tri-gram",
+    )
+
+    return parser
+
+
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    assert args.jit is False, "Support torchscript will be added later"
+
+    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> is 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)
+
+    model.to(device)
+
+    if params.avg_last_n > 0:
+        filenames = find_checkpoints(params.exp_dir)[: params.avg_last_n]
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+    elif params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+
+    model.eval()
+
+    model.to("cpu")
+    model.eval()
+
+    if params.jit:
+        logging.info("Using torch.jit.script")
+        model = torch.jit.script(model)
+        filename = params.exp_dir / "cpu_jit.pt"
+        model.save(str(filename))
+        logging.info(f"Saved to {filename}")
+    else:
+        logging.info("Not using torch.jit.script")
+        # Save it using a format so that it can be loaded
+        # by :func:`load_checkpoint`
+        filename = params.exp_dir / "pretrained.pt"
+        torch.save({"model": model.state_dict()}, str(filename))
+        logging.info(f"Saved to {filename}")
+
+
+if __name__ == "__main__":
+    formatter = (
+        "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+    )
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/spgispeech/ASR/pruned_transducer_stateless2/joiner.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/joiner.py

egs/spgispeech/ASR/pruned_transducer_stateless2/model.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/model.py

egs/spgispeech/ASR/pruned_transducer_stateless2/optim.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/optim.py

egs/spgispeech/ASR/pruned_transducer_stateless2/scaling.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/scaling.py

egs/spgispeech/ASR/shared

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