mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-05 15:14:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

use shuffled LibriSpeech cuts instead

Browse Source
This commit is contained in:
jinzr 2024-01-05 19:44:10 +08:00
parent f42258caf8
commit 9c29aaa61c
8 changed files with 26 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
egs/librispeech/ASR/conformer_ctc3/train.py
View File

@ -952,10 +952,10 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

7
egs/librispeech/ASR/conformer_mmi/train.py
View File

@ -771,10 +771,11 @@ def run(rank, world_size, args):
valid_ali = None valid_ali = None
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

6
egs/librispeech/ASR/lstm_transducer_stateless3/train.py
View File

@ -989,10 +989,10 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

6
egs/librispeech/ASR/pruned2_knowledge/train.py
View File

@ -817,10 +817,10 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

10
egs/librispeech/ASR/pruned_transducer_stateless7_streaming/train.py
View File

@ -1038,13 +1038,17 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
assert not (
params.mini_libri and params.full_libri
), f"Cannot set both mini-libri and full-libri flags to True, now mini-libri {params.mini_libri} and full-libri {params.full_libri}"
if params.mini_libri: if params.mini_libri:
train_cuts = librispeech.train_clean_5_cuts() train_cuts = librispeech.train_clean_5_cuts()
else: else:
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

6
egs/librispeech/ASR/pruned_transducer_stateless7_streaming_multi/train.py
View File

@ -1150,10 +1150,10 @@ def run(rank, world_size, args):
librispeech = LibriSpeech(manifest_dir=args.manifest_dir) librispeech = LibriSpeech(manifest_dir=args.manifest_dir)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
train_cuts = filter_short_and_long_utterances(train_cuts, sp) train_cuts = filter_short_and_long_utterances(train_cuts, sp)

6
egs/librispeech/ASR/zipformer/train.py
View File

@ -1174,10 +1174,10 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
train_cuts += librispeech.train_clean_360_cuts() train_cuts = librispeech.train_all_shuf_cuts()
train_cuts += librispeech.train_other_500_cuts() else:
train_cuts = librispeech.train_clean_100_cuts()
def remove_short_and_long_utt(c: Cut): def remove_short_and_long_utt(c: Cut):
# Keep only utterances with duration between 1 second and 20 seconds # Keep only utterances with duration between 1 second and 20 seconds

3
egs/librispeech/ASR/zipformer_mmi/train.py
View File

@ -990,10 +990,7 @@ def run(rank, world_size, args):
librispeech = LibriSpeechAsrDataModule(args) librispeech = LibriSpeechAsrDataModule(args)
# train_cuts = librispeech.train_clean_100_cuts()
if params.full_libri: if params.full_libri:
# train_cuts += librispeech.train_clean_360_cuts()
# train_cuts += librispeech.train_other_500_cuts()
train_cuts = librispeech.train_all_shuf_cuts() train_cuts = librispeech.train_all_shuf_cuts()
else: else:
train_cuts = librispeech.train_clean_100_cuts() train_cuts = librispeech.train_clean_100_cuts()