add speechio

egs/multi_zh-hans/ASR/prepare.sh

@@ -107,6 +107,9 @@ if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
  if [ -e ../../aishell4/ASR/data/fbank/.fbank.done ]; then
  cd data/fbank
  ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_test) .
+ ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_train_L) .
+ ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_train_M) .
+ ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_train_S) .
  ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_train_L.jsonl.gz) .
  ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_train_M.jsonl.gz) .
  ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_train_S.jsonl.gz) .

egs/speechio/ASR/local/compute_fbank_speechio.py

@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+# Copyright 2023 Xiaomi Corp. (authors: Fangjun Kuang
+# Zengrui Jin)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+"""
+This file computes fbank features of the ST-CMDS dataset.
+It looks for manifests in the directory data/manifests/stcmds.
+
+The generated fbank features are saved in data/fbank.
+"""
+
+import argparse
+import logging
+import os
+from pathlib import Path
+
+import torch
+from lhotse import CutSet, WhisperFbank, WhisperFbankConfig, Fbank, FbankConfig, LilcomChunkyWriter
+from lhotse.recipes.utils import read_manifests_if_cached
+
+from icefall.utils import get_executor, str2bool
+
+# Torch's multithreaded behavior needs to be disabled or
+# it wastes a lot of CPU and slow things down.
+# Do this outside of main() in case it needs to take effect
+# even when we are not invoking the main (e.g. when spawning subprocesses).
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+SPEECHIO_TESTSET_INDEX = 26 # Currently, from 0 - 26 test sets are open source.
+
+def compute_fbank_speechio(num_mel_bins: int = 80, speed_perturb: bool = False, fbank_dir: str = "data/fbank", whisper_fbank: bool = False):
+ src_dir = Path("data/manifests")
+ output_dir = Path(fbank_dir)
+ num_jobs = min(8, os.cpu_count())
+
+ dataset_parts = []
+ for i in range(SPEECHIO_TESTSET_INDEX + 1):
+ idx = f"{i}".zfill(2)
+ dataset_parts.append(f"SPEECHIO_ASR_ZH000{idx}")
+
+ prefix = "speechio"
+ suffix = "jsonl.gz"
+ manifests = read_manifests_if_cached(
+ dataset_parts=dataset_parts,
+ output_dir=src_dir,
+ prefix=prefix,
+ suffix=suffix,
+ )
+ assert manifests is not None
+
+ assert len(manifests) == len(dataset_parts), (
+ len(manifests),
+ len(dataset_parts),
+ list(manifests.keys()),
+ dataset_parts,
+ )
+
+ if whisper_fbank:
+ extractor = WhisperFbank(WhisperFbankConfig(num_filters=args.num_mel_bins, device='cuda'))
+ else:
+ extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
+
+ with get_executor() as ex: # Initialize the executor only once.
+ for partition, m in manifests.items():
+ if (output_dir / f"{prefix}_cuts_{partition}.{suffix}").is_file():
+ logging.info(f"{partition} already exists - skipping.")
+ continue
+ logging.info(f"Processing {partition}")
+ cut_set = CutSet.from_manifests(
+ recordings=m["recordings"],
+ supervisions=m["supervisions"],
+ )
+ cut_set = cut_set.compute_and_store_features(
+ extractor=extractor,
+ storage_path=f"{output_dir}/{prefix}_feats_{partition}",
+ # when an executor is specified, make more partitions
+ num_jobs=num_jobs if ex is None else 80,
+ executor=ex,
+ storage_type=LilcomChunkyWriter,
+ )
+ cut_set.to_file(output_dir / f"{prefix}_cuts_{partition}.{suffix}")
+
+
+def get_args():
+ parser = argparse.ArgumentParser()
+ parser.add_argument(
+ "--num-mel-bins",
+ type=int,
+ default=80,
+ help="""The number of mel bins for Fbank""",
+ )
+ parser.add_argument(
+ "--whisper-fbank",
+ type=str2bool,
+ default=False,
+ help="Use WhisperFbank instead of Fbank. Default: False.",
+ )
+ parser.add_argument(
+ "--fbank-dir",
+ type=Path,
+ default=Path("data/fbank"),
+ help="Path to directory with train/valid/test cuts.",
+ )
+ return parser.parse_args()
+
+
+if __name__ == "__main__":
+ formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+ logging.basicConfig(format=formatter, level=logging.INFO)
+
+ args = get_args()
+ compute_fbank_speechio(
+ num_mel_bins=args.num_mel_bins, fbank_dir=args.fbank_dir, whisper_fbank=args.whisper_fbank
+ )