mirror of
https://github.com/k2-fsa/icefall.git
synced 2025-08-09 18:12:19 +00:00
41b16d7838
* merge upstream * add SURT model and training * add libricss decoding * add chunk width randomization * decode SURT with libricss * initial commit for zipformer_ctc * remove unwanted changes * remove changes to other recipe * fix zipformer softlink * fix for JIT export * add missing file * fix symbolic links * update results * clean commit for SURT recipe * training libricss surt model * remove unwanted files * remove unwanted changes * remove changes in librispeech * change some files to symlinks * remove unwanted changes in utils * add export script * add README * minor fix in README * add assets for README * replace some files with symlinks * remove unused decoding methods * initial commit for SURT AMI recipe * fix symlink * add train + decode scripts * add missing symlink * change files to symlink * change file type
79 lines
3.1 KiB
Python
Executable File
79 lines
3.1 KiB
Python
Executable File
#!/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 adds source features as temporal arrays to the mixture manifests.
|
|
It looks for manifests in the directory data/manifests.
|
|
"""
|
|
import logging
|
|
from pathlib import Path
|
|
|
|
import numpy as np
|
|
from lhotse import CutSet, LilcomChunkyWriter, load_manifest, load_manifest_lazy
|
|
from tqdm import tqdm
|
|
|
|
|
|
def add_source_feats():
|
|
src_dir = Path("data/manifests")
|
|
output_dir = Path("data/fbank")
|
|
|
|
logging.info("Reading mixed cuts")
|
|
mixed_cuts_clean = load_manifest_lazy(src_dir / "cuts_train_clean.jsonl.gz")
|
|
mixed_cuts_reverb = load_manifest_lazy(src_dir / "cuts_train_reverb.jsonl.gz")
|
|
|
|
logging.info("Reading source cuts")
|
|
source_cuts = load_manifest(src_dir / "ihm_cuts_train_trimmed.jsonl.gz")
|
|
|
|
logging.info("Adding source features to the mixed cuts")
|
|
pbar = tqdm(total=len(mixed_cuts_clean), desc="Adding source features")
|
|
with CutSet.open_writer(
|
|
src_dir / "cuts_train_clean_sources.jsonl.gz"
|
|
) as cut_writer_clean, CutSet.open_writer(
|
|
src_dir / "cuts_train_reverb_sources.jsonl.gz"
|
|
) as cut_writer_reverb, LilcomChunkyWriter(
|
|
output_dir / "feats_train_clean_sources"
|
|
) as source_feat_writer:
|
|
for cut_clean, cut_reverb in zip(mixed_cuts_clean, mixed_cuts_reverb):
|
|
assert cut_reverb.id == cut_clean.id + "_rvb"
|
|
source_feats = []
|
|
source_feat_offsets = []
|
|
cur_offset = 0
|
|
for sup in sorted(
|
|
cut_clean.supervisions, key=lambda s: (s.start, s.speaker)
|
|
):
|
|
source_cut = source_cuts[sup.id]
|
|
source_feats.append(source_cut.load_features())
|
|
source_feat_offsets.append(cur_offset)
|
|
cur_offset += source_cut.num_frames
|
|
cut_clean.source_feats = source_feat_writer.store_array(
|
|
cut_clean.id, np.concatenate(source_feats, axis=0)
|
|
)
|
|
cut_clean.source_feat_offsets = source_feat_offsets
|
|
cut_writer_clean.write(cut_clean)
|
|
# Also write the reverb cut
|
|
cut_reverb.source_feats = cut_clean.source_feats
|
|
cut_reverb.source_feat_offsets = cut_clean.source_feat_offsets
|
|
cut_writer_reverb.write(cut_reverb)
|
|
pbar.update(1)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
|
|
logging.basicConfig(format=formatter, level=logging.INFO)
|
|
add_source_feats()
|