Commits · 15f83db09efa7ac21c2cc0b0ed9f5e816ac142ac · school / Capstone Design / 01 / AudioLM

Browse files

This project is mirrored from https://github.com/lucidrains/audiolm-pytorch.git. Pull mirroring failed May 17, 2023.
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer or owner.
Last successful update May 15, 2023.

Feb 17, 2023
- apply the same gradient penalty schedule for stft and multiscale discriminators · 15f83db0
  Ilya Borovik authored Feb 17, 2023
  
  15f83db0
- log all soundstream losses with specified frequency · 151edc9a
  Ilya Borovik authored Feb 17, 2023
  
  151edc9a
- make sure soundstream training runs with accelerate · 51cef900
  Phil Wang authored Feb 16, 2023
  
  View commits for tag 0.12.5 0.12.5
  
  51cef900
- instructions for distributed training · dea0e2fe
  Phil Wang authored Feb 16, 2023
  
  dea0e2fe
- fix ability to load either model or ema model from trainer saved object, for... · 13e9dc3b
  Phil Wang authored Feb 16, 2023
  View commits for tag 0.12.4 0.12.4
  
  13e9dc3b
Feb 16, 2023
- fix issue with trainers not saving optimizer states · 535fa4a1
  Phil Wang authored Feb 15, 2023
  
  View commits for tag 0.12.3 0.12.3
  
  535fa4a1
- add ability to use lion optimizer · 0d96c1af
  Phil Wang authored Feb 15, 2023
  
  View commits for tag 0.12.2 0.12.2
  
  0d96c1af
Feb 14, 2023
- increase odds transformers are trained successfully · aeb4962a
  Phil Wang authored Feb 13, 2023
  
  View commits for tag 0.12.1 0.12.1
  
  aeb4962a
- increase the odds someone trains soundstream successfully with a new attention... · 9145655b
  Phil Wang authored Feb 13, 2023
  View commits for tag 0.12.0 0.12.0
  
  9145655b
Feb 12, 2023
- readme · 653e6e91
  Phil Wang authored Feb 11, 2023
  
  653e6e91
Feb 11, 2023
- 0.11.19 · 50ced3b0
  Phil Wang authored Feb 10, 2023
  
  View commits for tag 0.11.19 0.11.19
  
  50ced3b0
- allow for disabling the prompt asking whether to clear previous results · 6b779a39
  Phil Wang authored Feb 10, 2023
  
  6b779a39
Feb 10, 2023
- 0.11.18 · c12d7a82
  Phil Wang authored Feb 09, 2023
  
  View commits for tag 0.11.18 0.11.18
  
  c12d7a82
- Merge pull request #85 from djqualia/patch-12 · b937b4d1
  Phil Wang authored Feb 09, 2023
  b937b4d1
- Remove print statement from debugging · 429ee43c
  djqualia authored Feb 09, 2023
  429ee43c
- need latest version of torch for complex stft discriminator distributed training · d700bc3b
  Phil Wang authored Feb 09, 2023
  
  View commits for tag 0.11.17 0.11.17
  
  d700bc3b
- fix a small bug thanks to @Lunariz · c8fe964a
  Phil Wang authored Feb 09, 2023
  
  View commits for tag 0.11.16 0.11.16
  
  c8fe964a
Feb 09, 2023
- incorporate all of @turian feedback · 6ed3aad4
  Phil Wang authored Feb 08, 2023
  
  View commits for tag 0.11.15 0.11.15
  
  6ed3aad4
- credit assignment · be547023
  Phil Wang authored Feb 08, 2023
  
  be547023
- allow maximal flexibility in normalizing, either when loading data, or doing... · 1831daed
  Phil Wang authored Feb 08, 2023
  View commits for tag 0.11.14 0.11.14
  
  1831daed
Feb 08, 2023
- stop using deprecated torch.norm, as suggested by @zhvng · 0863e897
  Phil Wang authored Feb 07, 2023
  
  0863e897
- fix l2 norm loss for log mel spectrogram differences in multi spectral recon... · c8b4b748
  Phil Wang authored Feb 07, 2023
  c8b4b748
- try one more time to fix complex stft discriminator network distributed training · cb442af8
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.11 0.11.11
  
  cb442af8
- make n_fft and n_mels complete configurable for soundstream for multi spectral recon loss · 92376359
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.9 0.11.9
  
  92376359
- make a guess to suppress a warning, but need expert guidance here · 6123c770
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.8 0.11.8
  
  6123c770
- average the loss across time and batch for multi spectral recon loss, seems a... · 8afeb585
  Phil Wang authored Feb 07, 2023
  View commits for tag 0.11.7 0.11.7
  
  8afeb585
- allow for ability to turn off multi spectral recon loss · 627d6fee
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.6 0.11.6
  
  627d6fee
- return the multi spectral recon loss · b6ff57d9
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.5 0.11.5
  
  b6ff57d9
- fix bug in complex stft discriminator output after switching to logits abs output · f84bf917
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.4 0.11.4
  
  f84bf917
- should use absolute value of complex value for final loss · e3a4e897
  Phil Wang authored Feb 07, 2023
  
  e3a4e897
- address https://github.com/lucidrains/audiolm-pytorch/issues/76 and give thanks to Andrey · 45bb030a
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.11.2 0.11.2
  
  45bb030a
Feb 07, 2023
- bring back complex stft discriminator, issue still not resolved · 965d6bbc
  Phil Wang authored Feb 06, 2023
  
  View commits for tag 0.11.1 0.11.1
  
  965d6bbc
- address equations 4 and 5 in the soundstream paper - "multi spectral recon... · 47871901
  Phil Wang authored Feb 06, 2023
  47871901
- bug fixes, thanks to @zhvng · c229f62a
  Phil Wang authored Feb 06, 2023
  
  View commits for tag 0.10.4 0.10.4
  
  c229f62a
- Merge pull request #75 from zhvng/minor_fixes · aebfc16b
  Phil Wang authored Feb 06, 2023
  aebfc16b
Feb 06, 2023
- small bug fixes · 018a37c9
  zhvng authored Feb 06, 2023
  
  018a37c9
Feb 04, 2023
- appreciation · 0ce7c656
  Phil Wang authored Feb 03, 2023
  
  0ce7c656
- 0.10.3 · 6a7ca66c
  Phil Wang authored Feb 03, 2023
  
  View commits for tag 0.10.3 0.10.3
  
  6a7ca66c
- Merge pull request #73 from zhvng/batch_unique · bf6dda9e
  Phil Wang authored Feb 03, 2023
  bf6dda9e
- batch unique consecutive in CoarseTransformerWrapper generate · ed630be2
  zhvng authored Feb 04, 2023
  
  ed630be2