JinZr
156af46a6e
applied text norm to valid & test cuts
2024-10-08 00:02:16 +08:00
JinZr
f0744877a6
minor updates
2024-10-07 23:32:03 +08:00
JinZr
32a7d2222d
minor updates to the scripts
2024-10-07 20:55:53 +08:00
JinZr
266e840475
fixed `+x
` permission
2024-10-07 16:10:13 +08:00
JinZr
b65eba2d15
fixed script for inference
2024-10-07 09:53:09 +08:00
zr_jin
93eedced80
Merge branch 'k2-fsa:master' into dev/asr/libritts
2024-10-06 10:14:05 -07:00
JinZr
01cc307664
fixed loss functions & scaling factors
2024-10-07 01:03:26 +08:00
JinZr
58f6562824
added scheduler w/ warmup
2024-10-06 19:07:07 +08:00
JinZr
d83ce89fca
fixed loss normalization & scaling factors
2024-10-06 15:55:49 +08:00
JinZr
e788bb4853
making MSD and MPD optional
2024-10-06 13:38:05 +08:00
JinZr
f9340cc5d7
refactored loss functions
2024-10-05 23:11:43 +08:00
Yu Lianjie
5c04c31292
fix open-commands path ( #1714 )
2024-09-20 12:38:52 +08:00
Fangjun Kuang
6f1abd832d
Fix exporting streaming zipformer models. ( #1755 )
2024-09-11 21:04:52 +08:00
Fangjun Kuang
329e34ac20
Test export onnx models for multi-zh-hans ( #1752 )
2024-09-10 19:29:19 +08:00
zr_jin
a394bf7474
fixed gss scripts for alimeeting
and ami
recipes ( #1749 )
2024-09-08 20:35:07 +08:00
zr_jin
65b8a6c730
fixed wrong default value for the alimeeting
recipe ( #1750 )
2024-09-08 20:34:49 +08:00
Fangjun Kuang
2ff0bb6a88
fix CI tests ( #1748 )
2024-09-08 17:42:55 +08:00
zr_jin
559c8a7160
fixed a typo in prepare.sh
for alimeeting recipes ( #1747 )
2024-09-08 17:10:17 +08:00
JinZr
1e65a976d0
added pesq and stoi for reconstruction performance evaluation
2024-09-08 15:37:06 +08:00
JinZr
c43977ea05
black formatted
2024-09-08 11:23:27 +08:00
JinZr
d45b400805
minor updates
2024-09-08 11:16:12 +08:00
JinZr
c236757674
* added script for inference
...
* minor updates
2024-09-07 23:33:52 +08:00
Fangjun Kuang
d4b4323699
Fix github actions CI tests ( #1744 )
2024-09-07 19:21:26 +08:00
Fangjun Kuang
f233ffa02a
Add docker images for torch 2.4.1 ( #1743 )
2024-09-07 18:17:04 +08:00
JinZr
12c7a16a5a
minor updates
2024-09-06 22:05:21 +08:00
JinZr
4483c6e700
tensorboard should work properly
2024-09-06 21:52:59 +08:00
JinZr
8da57a0449
black formatted
2024-09-06 21:21:58 +08:00
JinZr
0150961a33
minor fixes
2024-09-06 21:20:45 +08:00
JinZr
2e5055a847
minor updates
2024-09-06 18:16:18 +08:00
JinZr
91f7b1ce6f
sort of fixed DDP training issue
2024-09-06 18:07:50 +08:00
JinZr
2df992f98a
fixed a typo
2024-09-05 22:35:57 +08:00
JinZr
6e4a9ea85a
a little bit coarse commit
2024-09-05 22:30:07 +08:00
jinzr
dd82686a0f
init commit
2024-09-04 22:16:41 +08:00
Yifan Yang
cea0dbe7b1
fix gigaspeech_prepare.sh ( #1734 )
2024-08-28 12:15:01 +08:00
Xiaoyu Yang
a6c02a4d8c
zipformer BF16 training recipe ( #1700 )
...
Support Zipformer AMP +BF16 training
2024-08-23 09:42:22 +08:00
Yuekai Zhang
3b434fe83c
fix triton onnx export ( #1730 )
2024-08-23 09:33:46 +08:00
Xiaoyu Yang
3fc06cc2b9
Support AudioSet training with weighted sampler ( #1727 )
2024-08-22 15:27:25 +08:00
Xiaoyu Yang
5952972294
Keep the custom fields in libriheavy manifest ( #1719 )
2024-08-17 13:24:38 +08:00
Yifan Yang
6ac3343ce5
fix path in README.md ( #1722 )
2024-08-16 20:13:02 +08:00
Karel Vesely
1730fce688
split save_results()
-> save_asr_output()
+ save_wer_results()
( #1712 )
...
- the idea is to support `--skip-scoring` argument passed to a decoding
script
- created for Transducer decoding (non-streaming, streaming)
- it can be done also for CTC decoding... (not yet)
- also added `--label` for extra label in `streaming_decode.py`
- and also added `set_caching_enabled(True)`, which has no effect on
librispeech, but it leads to faster runtime on DBs with long
recordings (assuming `librispeech/zipformer` scripts are the
example scripts for other setups)
2024-08-13 23:02:14 +08:00
Fangjun Kuang
3b257dd5ae
Add docker images for torch 2.4 ( #1704 )
2024-07-25 16:46:24 +08:00
Yuekai Zhang
4af81af5a6
Update Zipformer-xl 700M Results on multi-hans-zh ( #1694 )
...
* add blank penalty
* update zipformer-xl results
* fix typo
2024-07-18 21:05:59 +08:00
zzasdf
11151415f3
fix error in accum_grad ( #1693 )
2024-07-17 17:47:43 +08:00
Fangjun Kuang
2e13298717
Refactor ctc greedy search. ( #1691 )
...
Use torch.unique_consecutive() to avoid reinventing the wheel.
2024-07-15 12:01:47 +08:00
Zengwei Yao
d47c078286
add decoding method of ctc-greedy-search in zipformer recipe ( #1690 )
2024-07-14 17:30:13 +08:00
Zengwei Yao
334beed2af
fix usages of returned losses after adding attention-decoder in zipformer ( #1689 )
2024-07-12 16:50:58 +08:00
Ziwei Li
f6febd658e
"-" replace "_" fix writing error ( #1687 )
2024-07-12 14:42:00 +08:00
Teo Wen Shen
19048e155b
Cast grad_scale in whiten to float ( #1663 )
...
* cast grad_scale in whiten to float
* fix cast in zipformer_lora
2024-07-11 15:12:30 +08:00
Yifan Yang
d65187ec52
Small fix ( #1686 )
2024-07-11 14:45:35 +08:00
Zengwei Yao
785f3f0bcf
Update RESULTS.md, adding results and model links of zipformer-small/medium CTC/AED models ( #1683 )
2024-07-09 20:04:47 +08:00