Code

Collection of minor changes#26

Code

Open

spinjo wants to merge 20 commits intomainheidelberg-hepml/lorentz-gatr:mainfrom cleanup3heidelberg-hepml/lorentz-gatr:cleanup3Copy head branch name to clipboard

+98-81

Commits

Commits on Nov 8, 2024

• Remove unused NANException

  spinjo

  committed
  129d8c5

Commits on Nov 10, 2024

• Add missing bracket in logs

  spinjo

  committed
  1e51fc6

Commits on Nov 18, 2024

• Remove schedulefree optimizer (didnt help), take Lion optimizer from 'pytorch_optimizer' instead of official repo for convenience (using the same seed, both implementations give the same loss up to…

  Show description for cef1b4c
  spinjo

  committed
  cef1b4c

• Add ADOPT optimizer (recent hype)

  spinjo

  committed
  38a04cb

Commits on Nov 19, 2024

• Make the code work with training.clip_grad_norm=None (turns off gradient clipping)

  spinjo

  committed
  f684317

• Updated config files to match main, removed all mentions of ScheduleFree and updated requirements to prevent package conflicts.

  

  victor breso

  authored and

  victor breso

  committed
  f409c16

• Merge branch 'main' into cleanup3

  spinjo

  authored
  dc83381

• Expanded the range of allowed numpy versions in the requirements

  

  victor breso

  authored and

  victor breso

  committed
  cba6778

Commits on Nov 20, 2024

• Speed up embedding code: Avoid for loop over batch indices

  spinjo

  committed
  0962998

• Speed up embedding code part 2: Also kill the 2nd for loop (very similar to the 1st)

  spinjo

  committed
  1d4dd0e

• Speed up embedding code part 3: Change the if conditions in case we have no global tokens or no spurions - the previous code crashed in this case

  spinjo

  committed
  6c69fcf

• Commentary on the numpy range expansion and opt_einsum latest version update in the requirements

  

  vbpla

  authored and

  vbpla

  committed
  42d8a06

Commits on Nov 22, 2024

• Added a disclaimer about compatibility issues of xformers on Mac

  

  vbpla

  authored and

  vbpla

  committed
  df3f753

Commits on Nov 23, 2024

• Remove mentions of equivariant join from mlp documentation

  spinjo

  committed
  e8ab29c

Commits on Nov 25, 2024

• Extend requirements to make data/collect_data.py work

  spinjo

  committed
  4dd7d25

• Fix inconsistencies in tests: remove gated_gelu_divide from tests; make tests_regression work without the BaseWrapper (which was removed a long time ago)

  spinjo

  committed
  16e2aed

Commits on Nov 27, 2024

• For amplitudes DSI, define inv_mean and inv_std with register_buffer to make them part of the state_dict

  spinjo

  committed
  03a3224

• Fix the last fix - setting buffers to None does not work

  spinjo

  committed
  9a7cacb

• Fixed a bug on the amplitude plot script and made the model store the grad norm even if we don't have clip_grad_norm

  

  vbpla

  authored and

  vbpla

  committed
  a559a36

Commits on Nov 28, 2024

• Shortened the clip_grad_norm calculation accounting for all options

  

  vbpla

  authored and

  vbpla

  committed
  d5c0bc7