Commits · 65d9c407bf30e9158beda741d469e90ab8cb8179 · 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.

Nov 20, 2022
- resolve issue with eos, maybe · 65d9c407
  Phil Wang authored Nov 19, 2022
  
  View commits for tag 0.0.57 0.0.57
  
  65d9c407
Nov 19, 2022
- cleanup coarse transformer · d2932c75
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.56 0.0.56
  
  d2932c75
- 0.0.55 · 94cef623
  Phil Wang authored Nov 18, 2022
  
  94cef623
- helper generate functions on trainer that forward to training wrapper · 6b1f93eb
  Phil Wang authored Nov 18, 2022
  
  6b1f93eb
- give semantic transformer a training wrapper too, refactor and do things right · ddb40ee3
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.54 0.0.54
  
  ddb40ee3
- preliminary training code for all attention nets completed, project management · 96800b81
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.53 0.0.53
  
  96800b81
- first pass at fine transformer trainer · 31edfade
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.52 0.0.52
  
  31edfade
- first pass at semantic transformer trainer · 28e12b7e
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.51 0.0.51
  
  28e12b7e
- wire up gradient penalty for all discriminators · 5739c8f0
  Phil Wang authored Nov 18, 2022
  
  View commits for tag 0.0.50 0.0.50
  
  5739c8f0
- remove final blocker around training code for coarse transformer, which... · 37d9efab
  Phil Wang authored Nov 18, 2022
  View commits for tag 0.0.49 0.0.49
  
  37d9efab
- allow for sound dataset to return multiple resample audio, since for training... · f760d083
  Phil Wang authored Nov 18, 2022
  View commits for tag 0.0.48 0.0.48
  
  f760d083
- soundstream trainer should auto resample the input audio to the correct... · 52cfee78
  Phil Wang authored Nov 18, 2022
  View commits for tag 0.0.47 0.0.47
  
  52cfee78
Nov 18, 2022
- product and project management · 6289eb0b
  Phil Wang authored Nov 17, 2022
  
  6289eb0b
- show import · 0dbb294d
  Phil Wang authored Nov 17, 2022
  
  0dbb294d
- allow for turning off cross entropy loss to preceding tokens (semantic,... · 84bc71d7
  Phil Wang authored Nov 17, 2022
  View commits for tag 0.0.46 0.0.46
  
  84bc71d7
- reminder · b0fbb485
  Phil Wang authored Nov 17, 2022
  
  b0fbb485
- always train with quantize dropout in soundstream · 490dbec1
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.45 0.0.45
  
  490dbec1
- for the researchers · fabccfd7
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.44 0.0.44
  
  fabccfd7
- not sure how to handle, but make a guess · acf0cff5
  Phil Wang authored Nov 17, 2022
  
  acf0cff5
- use a hack to fix an issue, refactor semantic later · 0b078214
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.43 0.0.43
  
  0b078214
- make it work end to end · 9c4c56d3
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.42 0.0.42
  
  9c4c56d3
- be able to reconstruct the coarse wav from coarse transformer to soundstream decoder · a17b189a
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.41 0.0.41
  
  a17b189a
- add some preliminary generation code for coarse transformer, primed on semantic tokens · 684da45c
  Phil Wang authored Nov 17, 2022
  
  View commits for tag 0.0.40 0.0.40
  
  684da45c
Nov 17, 2022
- another fix needed for variable lengthed coarse token ids · a18a0cab
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.39 0.0.39
  
  a18a0cab
- cleanup · a7cb08d8
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.38 0.0.38
  
  a7cb08d8
- prepare for variable lengthed coarse tokens in fine transformer as well 0.0.36 · bc626a64
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.36 0.0.36
  
  bc626a64
- just give coarse sequence in coarse transformer its own start token · 2c2d43d8
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.34 0.0.34
  
  2c2d43d8
- ensure semantic transformer generations always include an eos for each sequence · 5cf4e945
  Phil Wang authored Nov 16, 2022
  
  5cf4e945
- semantic token ids will have variable lengths because of unique consecutive,... · 70f02c5b
  Phil Wang authored Nov 16, 2022
  70f02c5b
- cleanup · c6bcd118
  Phil Wang authored Nov 16, 2022
  
  c6bcd118
- handle unique consecutive issue with generating semantic token ids · ea194173
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.31 0.0.31
  
  ea194173
- only encode raw text once when generating · 0806f62b
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.30 0.0.30
  
  0806f62b
- code for sampling semantic token ids from semantic transformer, input can be... · d9152df6
  Phil Wang authored Nov 16, 2022
  View commits for tag 0.0.29 0.0.29
  
  d9152df6
- prepare some sampling functions · 3ea0bcf4
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.28 0.0.28
  
  3ea0bcf4
- get some working scripts into readme · 6a0c7e73
  Phil Wang authored Nov 16, 2022
  
  View commits for tag 0.0.27 0.0.27
  
  6a0c7e73
- move soundstream to separate file · 416aaf7a
  Phil Wang authored Nov 16, 2022
  
  416aaf7a
Nov 16, 2022
- readme · f9fa68e7
  Phil Wang authored Nov 15, 2022
  
  f9fa68e7
- readme · 873f3c2e
  Phil Wang authored Nov 15, 2022
  
  873f3c2e
- oops · 95b8cd23
  Phil Wang authored Nov 15, 2022
  
  View commits for tag 0.0.25 0.0.25
  
  95b8cd23
- optionally allow for resampling directly within SoundDataset, if target_sample_khz specified · b6e5af78
  Phil Wang authored Nov 15, 2022
  
  b6e5af78