08c653c21a
fix onnx
2024-05-17 17:05:57 +08:00
9660550c12
add grad checkpoint
2024-05-02 22:57:18 +08:00
a472178dcc
add grad checkpoint
2024-05-02 10:39:33 +08:00
167a4f3a29
add grad checkpoint
2024-04-30 21:36:06 +08:00
7317cf8150
add grad checkpoint
2024-04-30 21:27:00 +08:00
7c43d87b9a
add grad checkpoint
2024-04-30 21:19:10 +08:00
1fb6a89af7
add grad checkpoint
2024-04-30 21:17:26 +08:00
fd9ff66e4a
add convnext
2024-04-30 21:06:28 +08:00
e9f7db9eb1
add convnext
2024-04-30 21:04:50 +08:00
76afe57e47
New generative model algorithm: Rectified Flow ( #184 )
...
Implements Rectified Flow in DiffSinger.
ref:
- https://github.com/gnobitab/RectifiedFlow
- https://github.com/yxlllc/ReFlow-VAE-SVC
---------
Co-authored-by: yqzhishen <yangqian_1015@icloud.com >
2024-04-17 22:57:54 +08:00
73c54c91ad
Change default mel_base to e
...
This completes stage 1 of log base migration plan
2024-04-17 21:54:02 +08:00
71b8fbe5b6
Fix path type
2024-04-17 21:49:15 +08:00
5ca6261deb
Refactor click options
2024-04-03 21:52:08 +08:00
f9580010f8
Add missing keys to generated dsconfig.yaml
2024-03-11 12:00:41 +08:00
1489030fc1
Add link to doc in template config
2024-03-11 01:00:45 +08:00
e17c428d42
Finish docs for v2.3.0 release ( #177 )
...
* Complete configuration schemas for tension and voicing
* Add docs for choosing variance parameters
2024-03-11 00:52:22 +08:00
cad4830f06
Remove opencpop.txt and opencpop-strict.txt
2024-03-10 20:12:31 +08:00
23478cbe61
Update default vocoder name in generated dsconfig.yaml
2024-03-09 22:48:11 +08:00
04bc269c9d
Disable clamping in DDPM due to loudness issues
2024-03-06 23:20:02 +08:00
4becc8a268
Prevent repeated GT vocoding
2024-03-05 01:11:05 -06:00
28563a1279
Support using e as the base number of the log mel-spectrogram ( #175 )
...
* Add `mel_base` and set `e` as default value
* Export dynamic `mel_base`
* Restore old configs
* Refactor mel extraction
* add mel-base check for ddsp vocoder
---------
Co-authored-by: yxlllc <llc1995@sina.com >
2024-03-04 22:39:36 +08:00
8066e7230c
Fix missing predict_* keys in generated dsconfig.yaml of variance models
2024-03-01 23:16:54 +08:00
f71160d48e
Fix unexpected access to NoneType when not using melody encoder
...
Resolve #173
2024-02-29 18:03:05 +08:00
f9791414ce
Fix KeyError on f0_embed_type when exporting ONNX
2024-02-28 16:00:58 +08:00
c49ecbd4d5
Add win_size and fft_size to dsconfig.yaml
2024-02-27 00:41:21 +08:00
115be5acaf
Export dsconfig.yaml and vocoder.yaml together with ONNX models
2024-02-26 23:19:51 +08:00
c16095bc55
Drop support for some old features and behaviors ( #172 )
...
* Drop support for discrete F0 embedding (reserved in ONNX exporter)
* Drop support for `interp_uv` configuration key
* Drop support for `train_set_name` and `valid_set_name` configuration keys
* Drop support for linear domain of random time stretching augmentation
* Drop support for `num_pad_tokens` configuration key
* Drop support for code backup before training
* Drop support for `ffn_padding` configuration key
* Drop support for random seeding
* Add placeholder to load old checkpoint
* Remove duplicate txt_embed layer (resuming may raise errors)
* Remove migration script and error message for transcriptions.txt
* Remove seed from batch shuffling
* Use direct access on some hparam keys
* Fix duplicate keys in YAML
* Rename `pndm_speedup` to `diff_speedup`
2024-02-25 20:57:02 +08:00
d34d61028c
Fix overlapped initializer names when exporting variance model
2024-02-25 20:47:44 +08:00
e5c79c0c20
New variance parameter: voicing ( #170 )
...
* Support voicing in acoustic and variance models (experimental)
* Fix KeyError
2024-02-24 01:17:42 +08:00
b19a515915
Set up new vocoder model in default config file
2024-02-20 17:55:19 +08:00
b8ba29975a
New variance parameter: tension ( #169 )
...
* Add `DeconstructedWaveform` class and base harmonic extraction algorithm (by yxlllc)
* Support tension in acoustic and variance models (experimental)
* Support Kth harmonic extraction
* Use square difference instead
* Support weighted multi-harmonic definition
* Unify all tension domains in one function
* Remove `hparams` from pitch extraction APIs
* Fix KeyError
2024-02-19 20:53:52 +08:00
ef4df5f4aa
Merge pull request #159 from hrukalive/get_strategy
...
Better than PL auto strategy selection
2024-02-03 20:23:05 +08:00
98ba675759
Formatting
2024-01-04 00:15:59 -06:00
5cf42a68e0
Better than PL
2024-01-03 23:43:12 -06:00
8c446b1743
Better strategy auto selection with kwargs override
2023-12-30 14:13:54 -06:00
0392c9aba3
Deal with **kwargs-like signatures
2023-12-11 12:46:24 +08:00
c2b8e75b6d
Sync image version
2023-12-08 00:08:40 +08:00
a1edc18a8d
Update architecture graph with melody encoder
2023-12-07 23:55:58 +08:00
246c286ee0
Enable shallow diffusion and AMP by default
2023-12-07 21:23:22 +08:00
3576076c89
Try to avoid NaN breathiness in very rare edge cases
2023-12-07 16:04:39 +08:00
a22ace1f8e
Add notice for TensorBoard issue with DDP
2023-12-04 23:35:02 +08:00
18f4cbdb21
Improve error message and add instructions for NSF-HiFiGAN exporting
2023-12-02 20:02:38 +08:00
07908243fa
Update f0_max to 1100 and sync with docs
2023-12-02 18:44:12 +08:00
baa48221d0
Adjust documentation about materials and pre-trained models
2023-12-01 20:51:14 +08:00
575ff7ba4c
Add an introduction of mutual influence between variance modules
2023-12-01 14:07:56 +08:00
f68df7bf94
Support specifying vocoder model path for exporting
2023-11-29 02:04:26 +08:00
6efadc483d
Fix interpolation error when all notes are rest
2023-11-27 22:30:52 +08:00
e5dfe804be
Fix AttributeError caused by is_relative_to in Python 3.8
2023-11-27 02:40:15 +08:00
5062ae284e
Remove misleading comments and keys
2023-11-25 11:48:27 +08:00
859ad2e3ec
Expose expr by default and support --freeze_glide
2023-11-25 11:24:35 +08:00
autumn