Hui Zhang
|
d94db47f78
|
fix rotary embeding
|
1 year ago |
|
Hui Zhang
|
0a5cc5556e
|
rope for streaming decoding
|
1 year ago |
|
Hui Zhang
|
b56fb85ca0
|
RoPE with position interpolation
|
1 year ago |
|
Hui Zhang
|
b91b1c9b08
|
support position interpolation for langer attention context windown length.
|
1 year ago |
|
Hui Zhang
|
55870ffbb3
|
fix bugs
|
1 year ago |
|
Hui Zhang
|
03e9ea9e52
|
add roformer
|
1 year ago |
夜雨飘零
|
31a4562ae8
|
[ASR]add squeezeformer model (#2755)
* add squeezeformer model
* change CodeStyle, test=asr
* change CodeStyle, test=asr
* fix subsample rate error, test=asr
* merge classes as required, test=asr
* change CodeStyle, test=asr
* fix missing code, test=asr
* split code to new file, test=asr
* remove rel_shift, test=asr
|
1 year ago |
YangZhou
|
42ff946007
|
[audio] mv paddlespeech/audio to paddleaudio (#2706)
* split paddlespeech/audio to paddleaudio.
* add sox io ,sox effect, kaldi native fbank to paddleaudio.
|
2 years ago |
tianhao zhang
|
1ea828c30e
|
fix attention val bug
|
2 years ago |
|
tianhao zhang
|
3d994f5c23
|
format wav2vec2 demo
|
2 years ago |
|
Hui Zhang
|
9277fcb8a8
|
fix attn can not train
|
2 years ago |
|
Hui Zhang
|
925abcca23
|
format
|
2 years ago |
|
Hui Zhang
|
feb27e2a84
|
fuse linear kv
|
2 years ago |
|
Hui Zhang
|
3adb20b468
|
eliminate shape and slice
|
2 years ago |
|
Hui Zhang
|
46088c0a16
|
elimiate attn transpose
|
2 years ago |
|
Hui Zhang
|
f9e3eaa024
|
transpose in matmul
|
2 years ago |
|
Hui Zhang
|
3d7ca93861
|
bool type slice
|
2 years ago |
|
Hui Zhang
|
c2c8a662b1
|
refactor reshape
|
2 years ago |
|
Hui Zhang
|
6de81d74d9
|
elimiete cast dtype for bool op
|
2 years ago |
|
Hui Zhang
|
b7388ce25a
|
eliminate useless unsqueese
|
2 years ago |
|
Hui Zhang
|
00b2c1c8fb
|
fix forward attention decoder caller
|
2 years ago |
|
Hui Zhang
|
9b66680ea4
|
Merge branch 'u2++_decoder' into u2pp_export
|
2 years ago |
|
tianhao zhang
|
0a95689461
|
support bitransformer decoder
|
2 years ago |
|
tianhao zhang
|
1a56a6e42b
|
add bitransformer decoder, test=asr
|
2 years ago |
|
Hui Zhang
|
53d6baff0b
|
format
|
2 years ago |
|
Hui Zhang
|
e3298c79ce
|
Merge branch 'develop' into u2_export
|
2 years ago |
|
Hui Zhang
|
8690a00bd8
|
add feature pipeline layer(cmvn, fbank), but to_static and jit.layer output is not equal
|
2 years ago |
|
Hui Zhang
|
63aeb747b0
|
more comment
|
2 years ago |
|
Hui Zhang
|
d638325c46
|
do not jit save forward; using slice for zeros([0,0,0,0]) tensor
|
2 years ago |
|
tianhao zhang
|
6745e9dd6b
|
fix dp init
|
2 years ago |
TianYuan
|
795eb7bd10
|
format paddlespeech with pre-commit (#2331)
|
2 years ago |
|
tianhao zhang
|
cdcb1a5316
|
s2t: fix encoder.py
|
2 years ago |
|
tianhao zhang
|
ed2819d7af
|
fix format test=asr
|
2 years ago |
|
tianhao zhang
|
ed80b0e2c3
|
fix multigpu training test=asr
|
2 years ago |
|
tianhao zhang
|
733ec7f2bc
|
fix conformer multi-gpu training test=asr
|
2 years ago |
|
Hui Zhang
|
ef37f73a01
|
fix cnn cache dy2st shape
|
2 years ago |
0x45f
|
e6ac8881f1
|
Fix comments
|
2 years ago |
|
0x45f
|
ac680aa783
|
Merge branch 'develop' of https://github.com/PaddlePaddle/PaddleSpeech into new_api
|
2 years ago |
|
0x45f
|
294b7b00bd
|
Supprot dy2st for conformer
|
2 years ago |
huangyuxin
|
75997d8277
|
merge
|
2 years ago |
|
Hui Zhang
|
e81849277e
|
att cache for streaming asr
|
2 years ago |
|
Hui Zhang
|
fb40602d94
|
refactor attention cache
|
2 years ago |
|
TianYuan
|
bc93bffbb4
|
replace logger.info with logger.debug in cli, change default log level to INFO
|
2 years ago |
|
Hui Zhang
|
dfdf450b22
|
fix #2013; and format
|
2 years ago |
|
huangyuxin
|
10819e0fa2
|
not install ctc on win, test=asr
|
2 years ago |
|
huangyuxin
|
b23bde8ec5
|
tensor.shape => paddle.shape(tensor)
|
2 years ago |
|
huangyuxin
|
4c09927f61
|
fix
|
2 years ago |
|
huangyuxin
|
e1888f9ae6
|
remove size,test=asr
|
2 years ago |
xiongxinlei
|
b1ef434983
|
update the max len compute method, test=doc
|
2 years ago |
|
xiongxinlei
|
0ea39f837b
|
add asr time limt configuration, test=doc
|
2 years ago |
