Commits · ed10de423a87182e935e259cca80f9ee04afca1f · 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.

Jan 21, 2023
- add learned multi-headed exponential moving average as an option for soundstream · ed10de42
  Phil Wang authored Jan 20, 2023
  
  View commits for tag 0.7.1 0.7.1
  
  ed10de42
- fix when turning off local attention for sound stream · 455def25
  Phil Wang authored Jan 20, 2023
  
  View commits for tag 0.6.3 0.6.3
  
  455def25
- switch to continuous positional bias, for the length extrapolation at inference time · d5890a24
  Phil Wang authored Jan 20, 2023
  
  View commits for tag 0.6.2 0.6.2
  
  d5890a24
- make dilation order in encoder and decoder configurable in soundstream · 8e3d1979
  Phil Wang authored Jan 20, 2023
  
  View commits for tag 0.6.1 0.6.1
  
  8e3d1979
Jan 20, 2023
- use the same activation placements as encodec, suggested by @inspirit · 36c39540
  Phil Wang authored Jan 20, 2023
  
  View commits for tag 0.6.0 0.6.0
  
  36c39540
Jan 19, 2023
- make sure to address the discriminator gradient issues uncovered by... · daeedb27
  Phil Wang authored Jan 18, 2023
  View commits for tag 0.5.1 0.5.1
  
  daeedb27
- make sure local attention blocks are followed by feedforward · ec642cd9
  Phil Wang authored Jan 18, 2023
  
  View commits for tag 0.5.0 0.5.0
  
  ec642cd9
- readme · 5a7fee75
  Phil Wang authored Jan 18, 2023
  
  5a7fee75
- add an extra assert to protect against empty audiofiles · 15391183
  Phil Wang authored Jan 18, 2023
  
  View commits for tag 0.4.8 0.4.8
  
  15391183
- readme · 9d992724
  Phil Wang authored Jan 18, 2023
  
  9d992724
- be able to set dataloader num workers on soundstream trainer · 58f13b44
  Phil Wang authored Jan 18, 2023
  
  View commits for tag 0.4.7 0.4.7
  
  58f13b44
- make sure one can also finely specify the max length of each target sample freq · 2f4407b4
  Phil Wang authored Jan 18, 2023
  
  View commits for tag 0.4.6 0.4.6
  
  2f4407b4
- fix order in which audio is resampled vs pad / curtailed · f7c26f3b
  Phil Wang authored Jan 18, 2023
  
  View commits for tag 0.4.5 0.4.5
  
  f7c26f3b
Jan 17, 2023
- also allow for single resample target hz for sounddataset · 7f788a62
  Phil Wang authored Jan 16, 2023
  
  View commits for tag 0.4.4 0.4.4
  
  7f788a62
- reduce confusion for target_sample_hz · 5d6cd874
  Phil Wang authored Jan 16, 2023
  
  View commits for tag 0.4.3 0.4.3
  
  5d6cd874
- fix an issue with resampling in sounddataset, thanks to @LWprogramming · 0f10c2e1
  Phil Wang authored Jan 16, 2023
  
  View commits for tag 0.4.2 0.4.2
  
  0f10c2e1
- just automatically handle t5_encode_text receiving a single string · 096f72a8
  Phil Wang authored Jan 16, 2023
  
  View commits for tag 0.4.1 0.4.1
  
  096f72a8
Jan 15, 2023
- allow for prefix-based self attention conditioning, as was done in VALL-E, and... · 654d4969
  Phil Wang authored Jan 14, 2023
  View commits for tag 0.4.0 0.4.0
  
  654d4969
Jan 12, 2023
- address two bugs, thanks to @apoorv2904 in issue... · 5058cc44
  Phil Wang authored Jan 11, 2023
  View commits for tag 0.3.6 0.3.6
  
  5058cc44
- need to access unwrapped soundstream in trainer · 4c3c3534
  Phil Wang authored Jan 11, 2023
  
  View commits for tag 0.3.5 0.3.5
  
  4c3c3534
Jan 11, 2023
- 0.3.4 · 36dc9e0d
  Phil Wang authored Jan 11, 2023
  
  View commits for tag 0.3.4 0.3.4
  
  36dc9e0d
- Merge pull request #46 from djqualia/patch-9 · 6ea8cf01
  Phil Wang authored Jan 11, 2023
  6ea8cf01
- Fix logging of recon_loss and small typo · b181beac
  djqualia authored Jan 10, 2023
  b181beac
- 0.3.3 · 8b70431f
  Phil Wang authored Jan 10, 2023
  
  View commits for tag 0.3.3 0.3.3
  
  8b70431f
- Merge pull request #44 from djqualia/patch-6 · c1d008c6
  Phil Wang authored Jan 10, 2023
  c1d008c6
- Add Accelerate init_tracker and log losses · c63b1362
  djqualia authored Jan 10, 2023
  c63b1362
- fix stft discriminator · 920a8a46
  Phil Wang authored Jan 10, 2023
  
  View commits for tag 0.3.2 0.3.2
  
  920a8a46
- fix stft discriminator loss, after fixing the distributed issue · c5ea4618
  Phil Wang authored Jan 10, 2023
  
  View commits for tag 0.3.1 0.3.1
  
  c5ea4618
Jan 10, 2023
- just do what encodec does by concatting real and imag after the stft transform... · 8259b0d0
  Phil Wang authored Jan 09, 2023
  View commits for tag 0.3.0 0.3.0
  
  8259b0d0
Jan 07, 2023
- update · 86b5e4a0
  Phil Wang authored Jan 06, 2023
  
  86b5e4a0
Jan 05, 2023
- fix issue with accelerator wrapped models when saving and loading · bcf77f91
  Phil Wang authored Jan 04, 2023
  
  View commits for tag 0.2.3 0.2.3
  
  bcf77f91
- actually save optim state for soundstream · e4d31ce4
  Phil Wang authored Jan 04, 2023
  
  View commits for tag 0.2.2 0.2.2
  
  e4d31ce4
- add helper save load methods to all the trainers, for saving optimizer states,... · 89a35002
  Phil Wang authored Jan 04, 2023
  View commits for tag 0.2.1 0.2.1
  
  89a35002
Dec 28, 2022
- bring the feedforward design from localvit to audiolm · deceb7a0
  Phil Wang authored Dec 27, 2022
  
  View commits for tag 0.2.0 0.2.0
  
  deceb7a0
Dec 16, 2022
- 0.1.22 · 94fe05d1
  Phil Wang authored Dec 15, 2022
  
  View commits for tag 0.1.22 0.1.22
  
  94fe05d1
- Merge pull request #37 from djqualia/patch-3 · 6178b7d6
  Phil Wang authored Dec 15, 2022
  6178b7d6
- Make the training sample output filename unique · 7f3b2d68
  djqualia authored Dec 15, 2022
  7f3b2d68
- 0.1.21 · d4e278b6
  Phil Wang authored Dec 15, 2022
  
  View commits for tag 0.1.21 0.1.21
  
  d4e278b6
- Merge pull request #34 from djqualia/patch-2 · b6c24794
  Phil Wang authored Dec 15, 2022
  b6c24794
Dec 15, 2022
- fix soundstream training · b0384e83
  Phil Wang authored Dec 14, 2022
  
  View commits for tag 0.1.20 0.1.20
  
  b0384e83