mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-09 17:14:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

minor updates

Browse Source
This commit is contained in:
marcoyang1998 2023-09-15 09:56:42 +08:00
parent cb85d4c337
commit 1bd6be03c1
2 changed files with 4 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
egs/libriheavy/ASR/zipformer_prompt_asr/decode_baseline.py
View File

@ -464,8 +464,8 @@ def decode_dataset(
) )
ref_words = ref_text.split() ref_words = ref_text.split()
this_batch.append((cut_id, ref_words, hyp_words)) this_batch.append((cut_id, ref_words, hyp_words))
if not params.use_ls_test_set: # if not params.use_ls_test_set:
results[name + " " + book_name].extend(this_batch) # results[name + " " + book_name].extend(this_batch)
results[name].extend(this_batch) results[name].extend(this_batch)
num_cuts += len(texts) num_cuts += len(texts)
@ -707,12 +707,6 @@ def main():
num_param = sum([p.numel() for p in model.parameters()]) num_param = sum([p.numel() for p in model.parameters()])
logging.info(f"Number of model parameters: {num_param}") logging.info(f"Number of model parameters: {num_param}")
def get_joint_last(texts: List[str], pre_texts: List[str]):
return {
"text": texts[-1],
"pre_text": pre_texts[-1]
}
# we need cut ids to display recognition results. # we need cut ids to display recognition results.
args.return_cuts = True args.return_cuts = True
libriheavy = LibriHeavyAsrDataModule(args) libriheavy = LibriHeavyAsrDataModule(args)
@ -722,7 +716,6 @@ def main():
ls_test_clean_cuts = libriheavy.librispeech_test_clean_cuts() ls_test_clean_cuts = libriheavy.librispeech_test_clean_cuts()
ls_test_other_cuts = libriheavy.librispeech_test_other_cuts() ls_test_other_cuts = libriheavy.librispeech_test_other_cuts()
long_audio_cuts = libriheavy.long_audio_cuts() long_audio_cuts = libriheavy.long_audio_cuts()
#test_clean_cuts = test_clean_cuts.filter(lambda c: "Brain Twister" not in c.text_path)
test_clean_dl = libriheavy.valid_dataloaders(test_clean_cuts,) test_clean_dl = libriheavy.valid_dataloaders(test_clean_cuts,)
test_other_dl = libriheavy.valid_dataloaders(test_other_cuts,) test_other_dl = libriheavy.valid_dataloaders(test_other_cuts,)

4
egs/libriheavy/ASR/zipformer_prompt_asr/decode_subformer_with_style.py
View File

@ -438,8 +438,8 @@ def decode_one_batch(
pre_texts = [t.lower() for t in pre_texts] pre_texts = [t.lower() for t in pre_texts]
if params.use_style_prompt: if params.use_style_prompt:
fixed_sentence = "Mixed-case English transcription, with punctuation. Actually, it is fully not related." fixed_sentence = "Mixed-case English transcription, with punctuation. Actually, it is fully not related. I'm hoping that this will lead to more accurate transcriptions."
style_texts = batch["supervisions"].get("style_text", [fixed_sentence for _ in range(batch_size)]) style_texts = [fixed_sentence for _ in range(batch_size)]
style_texts = [train_text_normalization(t) for t in style_texts] style_texts = [train_text_normalization(t) for t in style_texts]
else: else:
style_texts = ["" for _ in range(batch_size)] # use empty string style_texts = ["" for _ in range(batch_size)] # use empty string