added support for stcmds and primewords

This commit is contained in:
zr_jin 2023-07-17 13:28:59 +08:00
parent 0098d5eeb6
commit 34bc78f15c
3 changed files with 299 additions and 5 deletions

View File

@ -0,0 +1,114 @@
#!/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 Primewords dataset.
It looks for manifests in the directory data/manifests/primewords.
The generated fbank features are saved in data/fbank.
"""
import argparse
import logging
import os
from pathlib import Path
import torch
from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
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_thchs30(num_mel_bins: int = 80):
src_dir = Path("data/manifests/primewords")
output_dir = Path("data/fbank")
num_jobs = min(15, os.cpu_count())
dataset_parts = ("train",)
prefix = "primewords"
suffix = "jsonl.gz"
manifests = read_manifests_if_cached(
dataset_parts=dataset_parts,
output_dir=src_dir,
prefix=prefix,
suffix=suffix,
)
assert manifests is not None
assert len(manifests) == len(dataset_parts), (
len(manifests),
len(dataset_parts),
list(manifests.keys()),
dataset_parts,
)
extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
with get_executor() as ex: # Initialize the executor only once.
for partition, m in manifests.items():
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"],
)
if "train" in partition:
cut_set = (
cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
)
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""",
)
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_thchs30(num_mel_bins=args.num_mel_bins)

View File

@ -0,0 +1,118 @@
#!/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, Fbank, FbankConfig, LilcomChunkyWriter
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_thchs30(num_mel_bins: int = 80):
src_dir = Path("data/manifests/stcmds")
output_dir = Path("data/fbank")
num_jobs = min(15, os.cpu_count())
dataset_parts = (
"train",
"dev",
"test",
)
prefix = "stcmds"
suffix = "jsonl.gz"
manifests = read_manifests_if_cached(
dataset_parts=dataset_parts,
output_dir=src_dir,
prefix=prefix,
suffix=suffix,
)
assert manifests is not None
assert len(manifests) == len(dataset_parts), (
len(manifests),
len(dataset_parts),
list(manifests.keys()),
dataset_parts,
)
extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
with get_executor() as ex: # Initialize the executor only once.
for partition, m in manifests.items():
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"],
)
if "train" in partition:
cut_set = (
cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
)
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""",
)
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_thchs30(num_mel_bins=args.num_mel_bins)

View File

@ -101,14 +101,76 @@ if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
if [ -e ../../aishell/ASR/data/fbank/.aishell2.done ]; then
cd data/fbank
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_feats_train) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell_feats_dev) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell_feats_test) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell_cuts_train.jsonl.gz) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell_cuts_dev.jsonl.gz) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell_cuts_test.jsonl.gz) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_feats_dev) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_feats_test) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_cuts_train.jsonl.gz) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_cuts_dev.jsonl.gz) .
ln -svf $(realpath ../../../../aishell2/ASR/data/fbank/aishell2_cuts_test.jsonl.gz) .
cd ../..
else
log "Abort! Please run ../../aishell2/ASR/prepare.sh --stage 3 --stop-stage 3"
exit 1
fi
fi
log "Dataset: AISHELL-4"
if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
log "Stage 5: Prepare AISHELL-4"
if [ -e ../../aishell/ASR/data/fbank/.aishell4.done ]; then
cd data/fbank
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_train) .
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_dev) .
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_feats_test) .
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_train.jsonl.gz) .
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_dev.jsonl.gz) .
ln -svf $(realpath ../../../../aishell4/ASR/data/fbank/aishell4_cuts_test.jsonl.gz) .
cd ../..
else
log "Abort! Please run ../../aishell4/ASR/prepare.sh --stage 3 --stop-stage 3"
exit 1
fi
fi
log "Dataset: ST-CMDS"
if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
log "Stage 6: Prepare ST-CMDS"
if [ ! -f $dl_dir/stcmds/ST-CMDS-20170001_1-OS.tar.gz ]; then
log "Downloading ST-CMDS"
lhotse download stcmds $dl_dir/stcmds
fi
if [ ! -f data/manifests/.stcmds.done ]; then
mkdir -p data/manifests
lhotse prepare stcmds $dl_dir/stcmds data/manifests/stcmds
touch data/manifests/.stcmds.done
fi
if [ ! -f data/fbank/.stcmds.done ]; then
mkdir -p data/fbank
./local/compute_fbank_stcmds.py
touch data/fbank/.stcmds.done
fi
fi
log "Dataset: Primewords"
if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
log "Stage 6: Prepare Primewords"
if [ ! -f $dl_dir/primewords/primewords_md_2018_set1.tar.gz ]; then
log "Downloading Primewords"
lhotse download primewords $dl_dir/primewords
fi
if [ ! -f data/manifests/.stcmds.done ]; then
mkdir -p data/manifests
lhotse prepare stcmds $dl_dir/primewords data/manifests/primewords
touch data/manifests/.primewords.done
fi
if [ ! -f data/fbank/.primewords.done ]; then
mkdir -p data/fbank
./local/compute_fbank_primewords.py
touch data/fbank/.primewords.done
fi
fi