david.95
|
13a7fa9808
|
enable chinese words' pinyin specified in text of ssml formats, test=tts
|
2 years ago |
tianhao zhang
|
49c0cf9e31
|
format reference.md
|
2 years ago |
|
tianhao zhang
|
f29294153b
|
update reference.md and released_model.md
|
2 years ago |
Hui Zhang
|
7dc9cba3be
|
ctc prefix beam search for u2, test can run
|
2 years ago |
|
tianhao zhang
|
dbe8cee248
|
release wav2vec2ASR and wav2vec2.0 model, update Recent Update
|
2 years ago |
liangym
|
b76968e6d9
|
[tts] add mix finetune (#2525)
* updata readme, test=doc
* update yaml and readme, test=tts
* fix batch_size, test=tts
* add mix finetune, test=tts
* updata readme, test=tts
|
2 years ago |
|
Hui Zhang
|
3c3aa6b594
|
simple ctc prefix beam search compile ok
|
2 years ago |
|
Hui Zhang
|
bc1b6c2e7c
|
refactor ctc opts, extract decoder interface, add ctc beamsearch score
|
2 years ago |
|
Hui Zhang
|
5c8725e8cd
|
unify model opts; add attention rescore in decodable; rename ds2 ctc beam search
|
2 years ago |
|
Hui Zhang
|
6987751ff8
|
fix LogLikelihood and add AdvanceChunk
|
2 years ago |
|
Hui Zhang
|
1b9b1f5454
|
Merge pull request #2522 from Zth9730/u2pp_jit_export
[doc] update install introduction
|
2 years ago |
|
Hui Zhang
|
f1ca564731
|
Merge pull request #2518 from Zth9730/wav2vec2.0
[ASR] wav2vec2 ASR, pre-trained wav2vec2 based CTC for librispeech
|
2 years ago |
|
tianhao zhang
|
2ae94bd277
|
freeze wav2vec2=True, change loss report and update README.md
|
2 years ago |
|
tianhao zhang
|
e1a70ca1ed
|
update install introduction
|
2 years ago |
|
tianhao zhang
|
3d994f5c23
|
format wav2vec2 demo
|
2 years ago |
|
Hui Zhang
|
5cc874e1c3
|
u2 nnet get encoder out and align with py
|
2 years ago |
|
Hui Zhang
|
a75abc1828
|
fix u2 nnet out frames num
|
2 years ago |
|
Hui Zhang
|
cd1ced4ea0
|
add nnetout struct
|
2 years ago |
TianYuan
|
642232a577
|
Update install.md
|
2 years ago |
|
TianYuan
|
846434c05b
|
Update install_cn.md
|
2 years ago |
|
Hui Zhang
|
c6f9764ed6
|
Merge pull request #2510 from Zth9730/u2pp_jit_export
[s2t] use reverse_weight in decode.yaml
|
2 years ago |
|
tianhao zhang
|
7bee9d807f
|
format wav2vec2 demo
|
2 years ago |
|
tianhao zhang
|
19180d359d
|
format wav2vec2 demo
|
2 years ago |
|
tianhao zhang
|
6e429f0513
|
support wav2vec2ASR on librispeech
|
2 years ago |
|
Hui Zhang
|
290c23b9d7
|
add u2 nnet, u2 nnet main, codelab, and can compile
|
2 years ago |
|
tianhao zhang
|
e367242765
|
update dependency of paddle
|
2 years ago |
|
tianhao zhang
|
d2999ba21d
|
update install.md
|
2 years ago |
|
Hui Zhang
|
e1fc57deb1
|
add math and rename ds2 nnet
|
2 years ago |
|
Hui Zhang
|
75c578804d
|
using FetchContent_Declare for paddleinference
|
2 years ago |
|
Hui Zhang
|
b621b5b974
|
add math and macros
|
2 years ago |
|
Hui Zhang
|
532b620454
|
refactor speechx cmake
|
2 years ago |
|
tianhao zhang
|
5a66a14659
|
fix u2pp model version number
|
2 years ago |
|
tianhao zhang
|
cda440e6f0
|
use reverse_weight in decode.yaml
|
2 years ago |
Zth9730
|
c9b0c96b7b
|
Merge pull request #2502 from zh794390558/u2pp_export
[s2t] streaming conformer u2 and u2pp jit export
|
2 years ago |
|
Hui Zhang
|
c98b5dd173
|
fix masked_fill which will nan in trainning
|
2 years ago |
|
Hui Zhang
|
9277fcb8a8
|
fix attn can not train
|
2 years ago |
|
Hui Zhang
|
1f4f98b171
|
fix bug
|
2 years ago |
|
liangym
|
0359c3f6b5
|
Fix mix front (#2493)
* update mix frontend, test=tts
|
2 years ago |
|
Hui Zhang
|
e86337a423
|
fix bug
|
2 years ago |
|
Hui Zhang
|
abe22e56a4
|
paddele vertion for u2/u2pp export
|
2 years ago |
|
Hui Zhang
|
925abcca23
|
format
|
2 years ago |
|
Hui Zhang
|
2a75405e9a
|
Merge branch 'develop' into u2pp_export
|
2 years ago |
|
Hui Zhang
|
3ed24474d2
|
wenetspeech asr1 quant
|
2 years ago |
|
Hui Zhang
|
467cfd4e75
|
Merge pull request #2489 from Zth9730/u2++_server
[ASR] support u2pp based cli and server, optimiz code of u2pp decode (reversed_weight parameter)
|
2 years ago |
|
tianhao zhang
|
5b5167b586
|
support u2pp cli and server, optimiz code of u2pp decode, test=asr
|
2 years ago |
YangZhou
|
3507829a6d
|
Merge pull request #2464 from THUzyt21/Deploy-text-model-in-server
[Server]Deploy text model in server
|
2 years ago |
ZapBird
|
7a13b35fe6
|
BytesIO类型时，要保证切到初始位置，这样多次读取才能够正常。比如__call__函数。 (#2484)
* BytesIO类型时，要保证切到初始位置，这样多次读取才能够正常。比如__call__函数。
__call__函数的参数audio_file为BytesIO类型时执行到self.preprocess(model, audio_file)会报错，需要判断audio_file为BytesIO类型时执行audio_file.seek(0)。
|
2 years ago |
|
tianhao zhang
|
5bbe6e9897
|
support u2pp cli and server, optimiz code of u2pp decode, test=asr
|
2 years ago |
|
Hui Zhang
|
bdf876ea7b
|
Merge branch 'develop' into u2pp_export
|
2 years ago |
Zhao Yuting
|
304dc2603c
|
Update text_engine.py
|
2 years ago |
