Commits · b95e54b2250a7b775e2ce2f63d9daa6619025a3e · school / Capstone Design / 01 / MusicLM

Browse files

This project is mirrored from https://github.com/lucidrains/musiclm-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.

Mar 30, 2023
- contrastive loss needs both directions · b95e54b2
  Phil Wang authored Mar 29, 2023
  
  View commits for tag 0.1.2 0.1.2
  
  b95e54b2
Mar 02, 2023
- precompute audio and text layers to be used for hierarchical contrastive loss · cad7b494
  Phil Wang authored Mar 01, 2023
  
  View commits for tag 0.1.1 0.1.1
  
  cad7b494
- add the hierarchical contrastive loss from the ViCHA strategy in https://arxiv.org/abs/2208.13628 · 5f53d410
  Phil Wang authored Mar 01, 2023
  
  View commits for tag 0.1.0 0.1.0
  
  5f53d410
Feb 27, 2023
- project management · acee34a1
  Phil Wang authored Feb 26, 2023
  
  acee34a1
- allow for sampling multiple pieces of music and selecting top match with audio clip mulan · 1905a608
  Phil Wang authored Feb 26, 2023
  
  View commits for tag 0.0.28 0.0.28
  
  1905a608
Feb 17, 2023
- map location to cpu when loading state dict · 22fd63bb
  Phil Wang authored Feb 16, 2023
  
  View commits for tag 0.0.26 0.0.26
  
  22fd63bb
Feb 16, 2023
- fix save command not using the one on the trainer class, for saving optimizer state · c9b10bbc
  Phil Wang authored Feb 15, 2023
  
  View commits for tag 0.0.25 0.0.25
  
  c9b10bbc
Feb 15, 2023
- product management · 0247f2cc
  Phil Wang authored Feb 14, 2023
  
  0247f2cc
- 0.0.24 · 80cfebad
  Phil Wang authored Feb 14, 2023
  
  View commits for tag 0.0.24 0.0.24
  
  80cfebad
- fix a bug with getting the height width by patches in audio transformer, thanks to @ukemamaster · 50029726
  Phil Wang authored Feb 14, 2023
  
  50029726
Feb 14, 2023
- add qk rmsnorm · d514e2ff
  Phil Wang authored Feb 13, 2023
  
  View commits for tag 0.0.23 0.0.23
  
  d514e2ff
Feb 13, 2023
- avoid a division · 638b07b2
  Phil Wang authored Feb 12, 2023
  
  View commits for tag 0.0.22 0.0.22
  
  638b07b2
Feb 12, 2023
- repeat batch for relative positional bias only if doing patch dropout · d0cdbb25
  Phil Wang authored Feb 11, 2023
  
  View commits for tag 0.0.21 0.0.21
  
  d0cdbb25
- add best relative positional encoding for AST · 6b630672
  Phil Wang authored Feb 11, 2023
  
  View commits for tag 0.0.20 0.0.20
  
  6b630672
- last food for thought · 598aa341
  Phil Wang authored Feb 11, 2023
  
  598aa341
Feb 11, 2023
- readme · a77aa831
  Phil Wang authored Feb 10, 2023
  
  a77aa831
- add ability to skip prompt asking whether to clear previous results, by... · af8ff9a2
  Phil Wang authored Feb 10, 2023
  View commits for tag 0.0.19 0.0.19
  
  af8ff9a2
Feb 10, 2023
- need torch 1.12 for complex stft discriminator from audiolm · df17dcd0
  Phil Wang authored Feb 09, 2023
  
  View commits for tag 0.0.18 0.0.18
  
  df17dcd0
Feb 09, 2023
- in contrastive learning, make sure always the full batch · f2dc4628
  Phil Wang authored Feb 08, 2023
  
  View commits for tag 0.0.17 0.0.17
  
  f2dc4628
- batch size must be greater than 1 for contrastive learning · 0d1f2b3d
  Phil Wang authored Feb 08, 2023
  
  View commits for tag 0.0.16 0.0.16
  
  0d1f2b3d
Feb 08, 2023
- helpful assert · 6d45461a
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.0.15 0.0.15
  
  6d45461a
- forget about weight decay, and make sure to clip wavs by data_max_length · 846740c3
  Phil Wang authored Feb 07, 2023
  
  View commits for tag 0.0.14 0.0.14
  
  846740c3
Feb 07, 2023
- link to educator · 096a21b8
  Phil Wang authored Feb 06, 2023
  
  096a21b8
- position generating module is incompatible with patch dropout, new strategy · 83fc875e
  Phil Wang authored Feb 06, 2023
  
  83fc875e
- curtail to shortest, as no self attention masking strategy in audio spectrogram transformer yet · bb73ba7a
  Phil Wang authored Feb 06, 2023
  
  View commits for tag 0.0.12 0.0.12
  
  bb73ba7a
- get a mulan trainer out, remove wip · c138c935
  Phil Wang authored Feb 06, 2023
  
  View commits for tag 0.0.11 0.0.11
  
  c138c935
Feb 06, 2023
- add patch dropout · 7dcfc974
  Phil Wang authored Feb 05, 2023
  
  View commits for tag 0.0.10 0.0.10
  
  7dcfc974
Feb 04, 2023
- adopt dual patchnorm · 5aaee642
  Phil Wang authored Feb 03, 2023
  
  View commits for tag 0.0.9 0.0.9
  
  5aaee642
Feb 03, 2023
- ok it is working · 26c82682
  Phil Wang authored Feb 02, 2023
  
  View commits for tag 0.0.8 0.0.8
  
  26c82682
- product and project management · 2de8bb00
  Phil Wang authored Feb 02, 2023
  
  2de8bb00
- it is done · e96259b7
  Phil Wang authored Feb 02, 2023
  
  View commits for tag 0.0.7 0.0.7
  
  e96259b7
- use base class from audiolm-pytorch for type validation · 6a9e4262
  Phil Wang authored Feb 02, 2023
  
  View commits for tag 0.0.6 0.0.6
  
  6a9e4262
- override parameters, so if mulan embed quantizer is passed to audiolm... · 0503f7d1
  Phil Wang authored Feb 02, 2023
  View commits for tag 0.0.5 0.0.5
  
  0503f7d1
- working inference example · c40aee30
  Phil Wang authored Feb 02, 2023
  
  View commits for tag 0.0.4 0.0.4
  
  c40aee30
Feb 02, 2023
- readme · 8a66cce6
  Phil Wang authored Feb 01, 2023
  
  8a66cce6
- complete the residual vector quantization of the joint embedding music-text... · 99d1f044
  Phil Wang authored Feb 01, 2023
  99d1f044
- link to laion community · e81b2d47
  Phil Wang authored Feb 01, 2023
  
  e81b2d47
Feb 01, 2023
- todo · 888552af
  Phil Wang authored Jan 31, 2023
  
  888552af
- readme · 95595fb9
  Phil Wang authored Jan 31, 2023
  
  95595fb9
- add decoupled contrastive learning for mulan as an option · ff4c75f8
  Phil Wang authored Jan 31, 2023
  
  View commits for tag 0.0.2 0.0.2
  
  ff4c75f8