facebookresearch/optimizers

Issues

• sanity check with lingua

  #50 opened 2 months ago by SeunghyunSEO
  8

• AttributeError: module 'torch.distributed.tensor' has no attribute 'zeros'

  #48 opened 2 months ago by tsengalb99
  4

• ImportError in latest version

  #38 opened 2 months ago by francois-rozet
  0

• [bug]: `ImportError`: Cannot import name `DeviceMesh` from `torch.distributed`

  #36 opened 2 months ago by saforem2
  3

• Using Shampoo with Accelerate and FSDP

  #24 opened 2 months ago by kfirgoldberg
  3

• Failed to compute eigendecomposition

  #13 opened a year ago by aykamko
  5

• Empty params in FSDP cause issue

  #23 opened 3 months ago by odegeasslbc
  2

• ValueError when start_preconditioning_step value=0 and precondition_frequency=1

  #30 opened 3 months ago by bregaldo
  2

• ModuleNotFoundError: No module named 'optimizer_modules'

  #28 opened 3 months ago by francois-rozet
  1

• pip install-able

  #16 opened 4 months ago by windsornguyen
  3

• Fails from DeepSpeed

  #19 opened 4 months ago by catid-saronic
  1

• Typo in Algorithm 2 in the Distributed Shampoo paper: beta1 used instead of beta2

  #21 opened 4 months ago by jondeuce
  1

• Are eigh calls spread across gpus? or all on a single gpu?

  #17 opened 5 months ago by nikhilvyas
  1

• ModuleNotFoundError: No module named 'optimizer_modules', while trying to import the DistributedShampoo class

  #15 opened 5 months ago by ExpressGradient
  2

• using shampoo with distributed training

  #2 opened 3 years ago by maxmatical
  2

• Add setup.py

  #5 opened 2 years ago by LouisCastricato
  1

• resuming shampoo from checkpoint results in preconditioners with tensors on wrong device

  #3 opened 2 years ago by rwightman
  5

• Expectations

  #1 opened 3 years ago by avinashsai
  3