mirror of
https://github.com/k2-fsa/icefall.git
synced 2025-08-08 09:32:20 +00:00
5aafbb970e
* 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
105 lines
3.1 KiB
Python
Executable File
105 lines
3.1 KiB
Python
Executable File
#!/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()
|