mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 10:16:14 +00:00
Code Issues Packages Projects Releases Wiki Activity

update test set with wenetspeech test meeting

Browse Source
This commit is contained in:
root 2024-06-12 01:50:57 +00:00 committed by Yuekai Zhang
parent b26d3fa596
commit 3195a55ac7
4 changed files with 36 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
egs/speech_llm/ASR_LLM/whisper_llm_zh/decode.py
View File

@ -561,7 +561,8 @@ def main():
return True
# test_sets_cuts = multi_dataset.test_cuts()
test_sets_cuts = multi_dataset.aishell_test_cuts()
# test_sets_cuts = multi_dataset.aishell_test_cuts()
test_sets_cuts = multi_dataset.wenetspeech_test_meeting_cuts()
test_sets = test_sets_cuts.keys()
test_dls = [

4
egs/speech_llm/ASR_LLM/whisper_llm_zh/ds_config_zero1.json
View File

@ -19,14 +19,14 @@
"optimizer": {
"type": "Adam",
"params": {
"lr": 5e-4
"lr": 1e-4
}
},
"scheduler": {
"type": "WarmupLR",
"params": {
"warmup_min_lr": 0,
"warmup_max_lr": 5e-4,
"warmup_max_lr": 1e-4,
"warmup_num_steps": 100
}
},

41
egs/speech_llm/ASR_LLM/whisper_llm_zh/multi_dataset.py
View File

@ -230,20 +230,20 @@ class MultiDataset:
return {
"wenetspeech-meeting_test": wenetspeech_test_meeting_cuts,
# "aishell_test": aishell_test_cuts,
# "aishell_dev": aishell_dev_cuts,
# "ali-meeting_test": alimeeting_test_cuts,
# "ali-meeting_eval": alimeeting_eval_cuts,
# "aishell-4_test": aishell4_test_cuts,
# "aishell-2_test": aishell2_test_cuts,
# "aishell-2_dev": aishell2_dev_cuts,
# "magicdata_test": magicdata_test_cuts,
# "magicdata_dev": magicdata_dev_cuts,
# "kespeech-asr_test": kespeech_test_cuts,
# "kespeech-asr_dev_phase1": kespeech_dev_phase1_cuts,
# "kespeech-asr_dev_phase2": kespeech_dev_phase2_cuts,
# "wenetspeech-net_test": wenetspeech_test_net_cuts,
# "wenetspeech_dev": wenetspeech_dev_cuts,
"aishell_test": aishell_test_cuts,
"aishell_dev": aishell_dev_cuts,
"ali-meeting_test": alimeeting_test_cuts,
"ali-meeting_eval": alimeeting_eval_cuts,
"aishell-4_test": aishell4_test_cuts,
"aishell-2_test": aishell2_test_cuts,
"aishell-2_dev": aishell2_dev_cuts,
"magicdata_test": magicdata_test_cuts,
"magicdata_dev": magicdata_dev_cuts,
"kespeech-asr_test": kespeech_test_cuts,
"kespeech-asr_dev_phase1": kespeech_dev_phase1_cuts,
"kespeech-asr_dev_phase2": kespeech_dev_phase2_cuts,
"wenetspeech-net_test": wenetspeech_test_net_cuts,
"wenetspeech_dev": wenetspeech_dev_cuts,
}
def aishell_train_cuts(self) -> CutSet:
@ -317,4 +317,17 @@ class MultiDataset:
return {
"aishell2_test": aishell2_test_cuts,
}
def wenetspeech_test_meeting_cuts(self) -> CutSet:
logging.info("About to get multidataset test cuts")
# WeNetSpeech
logging.info("Loading WeNetSpeech set in lazy mode")
wenetspeech_test_meeting_cuts = load_manifest_lazy(
self.fbank_dir / "wenetspeech" / "cuts_TEST_MEETING.jsonl.gz"
)
return {
"wenetspeech-meeting_test": wenetspeech_test_meeting_cuts,
}

10
egs/speech_llm/ASR_LLM/whisper_llm_zh/train.py
View File

@ -823,14 +823,14 @@ def run(rank, world_size, args):
# an utterance duration distribution for your dataset to select
# the threshold
if c.duration < 1.0 or c.duration > 20.0:
logging.warning(
f"Exclude cut with ID {c.id} from training. Duration: {c.duration}"
)
# logging.warning(
# f"Exclude cut with ID {c.id} from training. Duration: {c.duration}"
# )
return False
return True
# train_cuts = multi_dataset.train_cuts()
train_cuts = multi_dataset.aishell_train_cuts()
train_cuts = multi_dataset.train_cuts()
# train_cuts = multi_dataset.aishell_train_cuts()
# train_cuts = multi_dataset.aishell2_train_cuts()
train_cuts = train_cuts.filter(remove_short_and_long_utt)