9e5222619f
Refactor: No preloading of BERT models to avoid unnecessary GPU VRAM consumption during training in the Web UI Since the BERT features of the dataset are pre-extracted by bert_gen.py, there is no need to load the BERT model at training time.
tsukumi
2024-03-12 18:22:34 +00:00
8baecb138a
Skip cuda test
litagin02
2024-03-13 11:54:24 +09:00
c37893a76b
Add version suffix dev0
litagin02
2024-03-13 11:46:56 +09:00
28fd715cb8
Restore gc empty_cache in train_ms_jp
litagin02
2024-03-13 10:25:22 +09:00
fedd019c04
Revert "Refactor: No preloading of BERT models to avoid unnecessary GPU VRAM consumption during training in the Web UI"
tsukumi
2024-03-12 19:01:45 +00:00
c4d6a8cdb7
Fix: large number of unnecessary files in built sdist Include in sdist only the minimum required files for style-bert-vits2 as a library.
tsukumi
2024-03-12 18:45:54 +00:00
07d246b98b
Refactor: run "hatch run style:fmt"
tsukumi
2024-03-12 18:27:08 +00:00
e8a76e547b
Refactor: No preloading of BERT models to avoid unnecessary GPU VRAM consumption during training in the Web UI Since the BERT features of the dataset are pre-extracted by bert_gen.py, there is no need to load the BERT model at training time.
tsukumi
2024-03-12 18:22:34 +00:00
483bc68d57
Refactor: unify invoke format of open() function
tsukumi
2024-03-12 18:00:51 +00:00
74af2c831c
Add: cache_dir and revision optional arguments to load_model() / load_tokenizer()
tsukumi
2024-03-12 17:45:10 +00:00
7f02b0f1d5
Refactor: TTSModelInfo changed from TypedDict to Pydantic model Pydantic models are more robust and properties can be accessed by dots.
tsukumi
2024-03-10 19:21:22 +00:00
859d940916
Fix: failed to start training
tsukumi
2024-03-10 19:15:11 +00:00
be265d42ed
Improve: switch pyworld to pyworld-prebuilt and enable it by default Prebuilt wheels for almost every OS/architecture (except musl) are now available on PyPI, eliminating the need for a build environment. ref: https://pypi.org/project/pyworld-prebuilt/
tsukumi
2024-03-10 16:51:58 +00:00
733a9d838d
Fix: clearly include Pydantic v2 in the dependencies The Pydantic models in the library are written for Pydantic v2 and will not work with Pydantic v1.
tsukumi
2024-03-10 15:38:08 +00:00
9c233630ef
Refactor: don't keep models in bert_feature module for each language BERT models and tokenizers are already stored and managed in the bert_models module and should not be stored here. In addition, since there may be situations where the user would like to use cpu instead of mps for inference when using it as a library, the automatic switching process to mps was removed.
tsukumi
2024-03-10 14:18:57 +00:00
00bf496325
Add: VSCode settings Enabling type checking with Pylance.
tsukumi
2024-03-10 13:51:56 +00:00
cdc47a98ce
Improve: test code
tsukumi
2024-03-10 11:10:06 +00:00
84b1dbe1b5
Fix: problem with test failures Style-Bert-VITS2 has been reported to not work with some PyTorch 2.2 series, but Python 3.12 is only supported in Torch >= 2.2, so Python 3.12 support is not provided for the time being
tsukumi
2024-03-10 10:48:21 +00:00
afff154da4
Add: test code for style-bert-vits2 as a library By executing "hatch run test:test", you can check whether the test passes in all Python 3.9 to 3.12 environments.
tsukumi
2024-03-10 04:18:39 +00:00
d2fd378b56
Refactor: rename Model / ModelHolder to TTSModel / TTSModelHolder for clarification and add comments to each method
tsukumi
2024-03-10 03:47:17 +00:00
b7d7c78203
Refactor: when pyopenjtalk_worker is called without initialization, continue processing without a worker When using style-bert-vits2 as a library, the requirement to be able to launch it in multiple processes may not be necessary. Also, if the library is embedded and exe-ed using PyInstaller or similar, it is difficult to make pyopenjtalk_worker run in a separate process. Therefore, we changed it so that the worker is used only when it is explicitly initialized.
tsukumi
2024-03-10 03:04:27 +00:00
a79b1910fb
Add: use hatch to build style-bert-vits2 as a library
tsukumi
2024-03-09 18:38:16 +00:00
0a74fe521b
Merge branch 'dev' of github.com:litagin02/Style-Bert-VITS2 into dev
tsukumi
2024-03-09 18:12:56 +00:00
1d320915e0
Refactor: make tts_model.py independent of style_gen.py
tsukumi
2024-03-09 18:10:25 +00:00
96d22102f3
Refactor: separate adjust_voice() function from tts_model.py
tsukumi
2024-03-09 17:45:56 +00:00
61e2a1deae
Refactor: add type hints to style_bert_vits2.models.utils
tsukumi
2024-03-09 17:26:42 +00:00
98ab8e7978
Refactor: separate module for utilities related to loading/saving checkpoints and safetensors
tsukumi
2024-03-09 15:58:57 +00:00
c594f7ea7a
Refactor: change execution location of pyopenjtalk.initialize() Considering library design, this function with many side effects should not be executed in a library.
tsukumi
2024-03-09 00:26:51 +00:00
8feef04cef
Refactor: add type hints to models.py / models_jp_extra.py I didn't add docstring because it is very technical code and I don't understand what is being implemented.
tsukumi
2024-03-08 23:52:44 +00:00
e1fad54c99
Refactor: add type hints to attentions.py / modules.py / transforms.py I didn't add docstring because it is very technical code and I don't understand what is being implemented.
tsukumi
2024-03-08 22:58:40 +00:00
717ba7925f
Fix: app.py cannot be closed with Ctrl+C
tsukumi
2024-03-08 22:27:21 +00:00
b98fecb46d
Add: --host/--port option to app.py to allow specifying listening host/port
tsukumi
2024-03-08 22:14:39 +00:00
d22a11ebb2
Fix: a bug that prevented speech synthesis in app.py
tsukumi
2024-03-08 22:09:47 +00:00
30ea08d6ea
Fix: forgot to write pyopenjtalk.initialize()
tsukumi
2024-03-08 18:22:40 +00:00
6b19eac61e
Merge branch 'dev' of github.com:litagin02/Style-Bert-VITS2 into dev
tsukumi
2024-03-08 16:02:22 +00:00
a84783a6cc
Refactor: replace utils.HParams with HyperParameters Pydantic model HyperParameters is largely a drop-in replacement for utils.HParams, which ensures type safety for hyper-parameters.
tsukumi
2024-03-08 15:52:37 +00:00
a0b5cd3d1b
Extended timeout in client side
kale4eat
2024-03-09 00:23:23 +09:00
df68771651
Refactor: split style_bert_vits2.nlp.chinese package Configured so that the same public function is exported from the module with the same name for each language.
tsukumi
2024-03-08 09:04:40 +00:00
5de4884075
Fix: import error pyopenjtalk_worker.initialize() has the side effect of starting another process and should not be executed automatically on import.
tsukumi
2024-03-08 07:31:33 +00:00
fac4f9a8ab
Refactor: rename text_processing to nlp "text_processing" is clearer, but the import statement is longer. "nlp" is shorter and makes it clear that it is natural language processing.
tsukumi
2024-03-08 06:20:44 +00:00
litagin02