Issues

• Need POC for Fairscale PyTorch Ecosystem Review

  facebookresearch/fairscale#1205 · Filter by author michellearoth opened on Sep 30, 2026

• Why autocast in checkpoint_activation uses default dtype?

  facebookresearch/fairscale#1199 · Filter by author psolt-leonardo opened on Aug 4, 2025

• Support free-threading Python wheel distributions

  facebookresearch/fairscale#1197 · Filter by author andfoy opened on Apr 21, 2025

• Error running large model of esm15B using FSDP

  facebookresearch/fairscale#1194 · Filter by author BKRBH opened on Feb 19, 2025

• AttributeError: 'FlatParameter' object has no attribute '_full_param_padded'

  facebookresearch/fairscale#1193 · Filter by author hui-li-xf opened on Jan 19, 2025

• When mixed precision input contains grad tensor, FSDP cast it with no grad

  facebookresearch/fairscale#1191 · Filter by author kamwoh opened on Dec 8, 2024

• Hi, Groups division may be incorrect in initialize() in fairscale/nn/model_parallel/initialize.py

  facebookresearch/fairscale#1189 · Filter by author Youngluc opened on Aug 30, 2024

• Raising assert param.grad is not None when finetuning LoRA.

  facebookresearch/fairscale#1188 · Filter by author HashimotoPatrickMu opened on Jun 26, 2024
  1

• [question] Different training between DDP & Sharded DDP

  facebookresearch/fairscale#1172 · Filter by author kwohlfahrt opened on Mar 29, 2024

• Example of MOE

  facebookresearch/fairscale#1165 · Filter by author Juanhui28 opened on Feb 28, 2024
  1

• How can I use torchrun + model parallelism + FSDP

  facebookresearch/fairscale#1155 · Filter by author HackGiter opened on Dec 8, 2023
  1

• It is dangerous to using default non_block=True.

  facebookresearch/fairscale#1146 · Filter by author heshenghuan opened on Oct 27, 2023

• assert self.has_full_params

  facebookresearch/fairscale#1134 · Filter by author pokameng opened on Sep 11, 2023
  4

• Why ShardedDDP and OSS are slower than Vanilla DDP

  facebookresearch/fairscale#1131 · Filter by author powermano opened on Aug 18, 2023

• pip install failed

  facebookresearch/fairscale#1130 · Filter by author dogxxxxx opened on Aug 17, 2023

• Error with nested models "Caffe2 uses a lazy allocation..."

  facebookresearch/fairscale#1129 · Filter by author Emanuele97x opened on Jul 26, 2023

• [bug] pip package 0.4.13 fails to build wheel

  facebookresearch/fairscale#1128 · Filter by author project-tuva opened on Jul 17, 2023

• Error Freezing Weights

  facebookresearch/fairscale#1126 · Filter by author mostafaelhoushi opened on Jun 5, 2023

• Compatibility with Pytorch 2.0; failing test test_gradient_value

  facebookresearch/fairscale#1124 · Filter by author h-vetinari opened on May 13, 2023
  4

• Can exclude some layer parameter not to shard?

  facebookresearch/fairscale#1123 · Filter by author robotcator opened on Apr 24, 2023
  5

• FSDP cannot consolidate optimizer state dict with flatten params is False

  facebookresearch/fairscale#1100 · Filter by author ShenglongZ opened on Dec 13, 2022
  3

• clip_grad_norm_ from fairscale downcasts to bf16 before all reduce

  facebookresearch/fairscale#1092 · Filter by author glample opened on Nov 2, 2022
  3

• Can't load optimizer state due to state_steps

  facebookresearch/fairscale#1083 · Filter by author rowhanm opened on Sep 26, 2022
  10

• Running stats with gradient checkpointing

  facebookresearch/fairscale#1035 · Filter by author vovaf709 opened on Jul 20, 2022
  8

• How to exclude some operations in checkpoint wrapper?

  facebookresearch/fairscale#1014 · Filter by author kenmbkr opened on Jun 22, 2022
  1