mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 18:24:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

misc. update

Browse Source
This commit is contained in:
jinzr 2024-04-08 17:19:42 +08:00
parent b9d34fb9d9
commit 05e48ca880
3 changed files with 10 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
egs/commonvoice/ASR/RESULTS.md
View File

@ -14,7 +14,7 @@ The best CER, for CommonVoice 16.1 (cv-corpus-16.1-2023-12-06/zh-HK) is below:
| modified_beam_search | 0.98 | 1.11 | --epoch 24 --avg 5 |
| fast_beam_search | 1.08 | 1.27 | --epoch 24 --avg 5 |
When doing the cross-corpus validation on MDCC (w/o blank penalty),
When doing the cross-corpus validation on [MDCC](https://arxiv.org/abs/2201.02419) (w/o blank penalty),
the best CER is below:
| | Dev | Test | Note |
@ -23,7 +23,7 @@ the best CER is below:
| modified_beam_search | 39.73 | 39.19| --epoch 24 --avg 5 |
| fast_beam_search | 42.14 | 41.98| --epoch 24 --avg 5 |
When doing the cross-corpus validation on MDCC (with blank penalty set to 2.2),
When doing the cross-corpus validation on [MDCC](https://arxiv.org/abs/2201.02419) (with blank penalty set to 2.2),
the best CER is below:
| | Dev | Test | Note |
@ -68,14 +68,16 @@ Detailed experimental results and pre-trained model are available at:
<https://huggingface.co/zrjin/icefall-asr-commonvoice-zh-HK-zipformer-2024-03-20>
### GigaSpeech BPE training results (Pruned Stateless Transducer 7)
### CommonVoice English (en) BPE training results (Pruned Stateless Transducer 7)
#### [pruned_transducer_stateless7](./pruned_transducer_stateless7)
See #997 for more details.
See #997 for more details.
Number of model parameters: 70369391, i.e., 70.37 M
Note that the result is obtained using GigaSpeech transcript trained BPE model
The best WER, as of 2023-04-17, for Common Voice English 13.0 (cv-corpus-13.0-2023-03-09/en) is below:
Results are:

1
egs/commonvoice/ASR/local/compute_fbank_commonvoice_splits.py
View File

@ -47,6 +47,7 @@ def get_args():
"--subset",
type=str,
default="train",
choices=["train", "validated", "invalidated"],
help="""Dataset parts to compute fbank. """,
)

6
egs/commonvoice/ASR/local/word_segment_yue.py
View File

@ -130,11 +130,11 @@ if __name__ == "__main__":
norm_lines = [normalize_text(line, lang) for line in lines]
text_words_segments = get_word_segments(norm_lines)
with open(output_dir / "transcript_words.txt", "w+", encoding="utf-8") as f:
with open(output_dir / "transcript_words.txt", "w", encoding="utf-8") as f:
f.writelines(text_words_segments)
words = get_words(text_words_segments)[1:] # remove "\n" from words
with open(output_dir / "words_no_ids.txt", "w+", encoding="utf-8") as f:
with open(output_dir / "words_no_ids.txt", "w", encoding="utf-8") as f:
f.writelines([word + "\n" for word in sorted(words)])
words = (
@ -143,5 +143,5 @@ if __name__ == "__main__":
+ ["#0", "<s>", "<\s>"]
)
with open(output_dir / "words.txt", "w+", encoding="utf-8") as f:
with open(output_dir / "words.txt", "w", encoding="utf-8") as f:
f.writelines([f"{word} {i}\n" for i, word in enumerate(words)])