源文雨
f4121844e6
doc: sync to latest grammar ( #427 )
2024-06-24 22:11:56 +09:00
源文雨
8235a46711
optimize: specify infer parameters ( #422 )
...
and move infer into core
2024-06-24 19:45:14 +09:00
源文雨
a9cb840538
doc: optimize README ( #421 )
2024-06-24 17:33:13 +09:00
源文雨
c8cb6bd327
feat: add normalizer interface & move instances out ( #420 )
2024-06-24 17:28:14 +09:00
源文雨
b62e0dc3c5
optimize(utils): move custom processors into model ( #419 )
2024-06-24 16:27:15 +09:00
源文雨
e0a9e7e043
fix(chat): model unload memory release ( #418 )
2024-06-24 16:19:05 +09:00
源文雨
21c8ecc960
optimize(chat): move gpt&decoders out of dict ( #416 )
...
for a clear type definition
2024-06-24 15:55:33 +09:00
Billy Cao
109376c326
fix(webui): disable torch.compile on Windows ( #413 )
2024-06-24 15:08:33 +09:00
源文雨
21a216f6d2
Merge pull request #414 from aliencaocao/fix-webui-demo
...
Add init.py to tools and web to fix launching webui demo
2024-06-24 09:17:50 +09:00
源文雨
46b007e140
Merge pull request #415 from 2noise/optimzie
...
optimize: all
2024-06-24 09:14:57 +09:00
源文雨
51c2118b76
fix(log): utils log cannot display
2024-06-23 23:45:50 +09:00
源文雨
d93bc19f41
optimize(dl): use mmap to hash
...
https://stackoverflow.com/a/67412295
2024-06-23 23:37:38 +09:00
源文雨
1b16feec2f
chore(core): add more types
2024-06-23 22:45:20 +09:00
源文雨
5d20028316
optimize(utils): jit HomophonesReplacer
2024-06-23 22:39:11 +09:00
Billy Cao
724f524629
Add init.py to tools and web to fix launching webui demo
2024-06-23 20:33:12 +08:00
源文雨
79684e2426
feat(webui): make Generate button be primary
2024-06-23 21:26:23 +09:00
源文雨
93ef96b173
feat: apply TorchSeedContext
2024-06-23 21:22:23 +09:00
源文雨
fef7084d1c
feat(webui): add reload button & dvae coef
2024-06-23 18:04:56 +09:00
源文雨
78ee6965b7
chore(webui): unify display language
2024-06-23 17:21:23 +09:00
源文雨
4dd1f88073
optimize: core & model
...
### all
- reduce redundant code
- add `_` to some inner calls
### core
- rename `check_model` to `has_loaded`
- new api `download_models`
- move vocos out of models dict (further target: remove the dict `pretrain_models`)
- new api `unload`
- new api `decode_to_wavs`
### model
- fix grad mul issue in dvae
- optimize progress bar display
- optimize tensor operations
2024-06-23 17:17:33 +09:00
源文雨
2eb97d212f
optimize: log printing ( #411 )
2024-06-23 01:27:49 +09:00
源文雨
1430c95a22
feat(ipynb): apply logger to notebook ( #410 )
2024-06-23 00:17:19 +09:00
源文雨
7fecdb8657
fix(dvae): given NumPy array is not writable ( #409 )
2024-06-23 00:06:54 +09:00
源文雨
98bf2c32c0
fix(core): local download path ( #408 )
2024-06-23 00:01:00 +09:00
源文雨
b5fd820a53
optimize(gpt): apply type definitions ( #407 )
2024-06-22 23:50:29 +09:00
源文雨
2ba7c6c138
feat(dvae): expose coef for customize ( #405 )
...
and unify coef of dvae & decoder
2024-06-22 19:59:53 +09:00
源文雨
b4c3cff4b2
fix(dvae): dvae _embed permute mismatch ( #403 )
...
when use_decoder=False
introduced in #383
maybe related to #400
2024-06-22 16:09:51 +09:00
zhipeng and 源文雨
a41ce6320b
feat(cmd): add multiple texts ( #366 )
...
Co-authored-by: 源文雨 <41315874+fumiama@users.noreply.github.com >
2024-06-22 15:16:04 +09:00
源文雨
12cc9f6686
chore: unify file names in utils ( #402 )
2024-06-22 15:03:13 +09:00
源文雨
3e0d87bca0
fix(tools): log formatting ( #401 )
...
- introduce numba
- move audio normalization functions to tools/audio
- remove IPython from tools/cmd
2024-06-22 14:59:06 +09:00
源文雨
f809de7289
fix(req): restrict normalizer' installation to linux ( #399 )
2024-06-22 01:54:21 +09:00
Faych Chen and neverbiasu
b29544ebb0
doc: revise the Load Models part of examples/ipynb ( #397 )
...
Co-authored-by: neverbiasu <neverbiasu@outlook.com >
2024-06-22 01:50:10 +09:00
源文雨
a851b9a9fe
optimize: log & webui ( #398 )
...
- move log definition out of ChatTTS
- apply colorful log level
- optimize webui logic
- split webui into 2 files for clear reading
2024-06-22 01:47:56 +09:00
libukai
e58fe48d2e
doc: sync CN README ( #394 )
2024-06-21 18:11:13 +09:00
源文雨
48be41922a
fix(examples): colab infer ( #393 )
2024-06-21 17:45:33 +09:00
源文雨
f9e6d2d9d3
fix(examples): ipnyb infer ( #392 )
2024-06-21 17:36:43 +09:00
源文雨
f17bd31327
fix: model loading check ( #391 )
2024-06-21 16:57:44 +09:00
源文雨
0d6621dee5
doc: release requirements limit ( #390 )
2024-06-21 16:47:35 +09:00
刘悦
e06fd5bd79
feat(webui): support stream infer ( #380 )
2024-06-21 16:30:24 +09:00
源文雨
d57dde191e
fix: mps memory leakage by moving gpt&emb to cpu ( #389 )
...
see more at
#261
https://github.com/pytorch/pytorch/issues/91368
https://stackoverflow.com/questions/74902903/pytorch-memory-leak-reference-cycle-in-for-loop
2024-06-21 16:29:10 +09:00
ermian
c42852ea38
fix: GPU memory check threshold ( #387 )
2024-06-21 16:18:43 +09:00
ermian
56227f466d
fix: homophones replacer ( #386 )
...
and improve output clarity
- 原代码在应用字符映射后错误地使用了变量t而不是text[i],会导致前一步 apply_character_map 的结果被覆盖。现已更正此问题。
- 同音字替换的输出已改进为仅显示替换的字符，使结果更简洁。
2024-06-21 16:17:59 +09:00
源文雨
ace4d0ca23
optimize(dvae): remove einops ( #383 )
2024-06-21 13:49:36 +09:00
源文雨
e6412b1c8b
fix(ipy): missing envs ( #379 )
2024-06-21 00:16:34 +09:00
asamaayako
153ccde9bb
fix: Windows not yet supported for torch.compile ( #377 )
2024-06-20 22:35:30 +09:00
LeoN0425
654b0f291b
optimize: mps audio quality by contiguous scores ( #373 )
...
Ensuring logits scores are contiguous resolved the bug on MPS
2024-06-20 17:53:19 +09:00
源文雨
3a936ea2f5
doc: optimize README ( #374 )
2024-06-20 17:45:21 +09:00
源文雨
4cd92fa721
doc: optimize README ( #368 )
2024-06-20 01:11:47 +09:00
源文雨
001d164049
doc: optimize README ( #367 )
2024-06-20 01:06:49 +09:00
libukai
a75bb6920a
doc: fix the readme link ( #365 )
2024-06-19 23:03:11 +09:00
源文雨