Search results

Need to specify minimal rust version to build

Status: Open.

#330 In meta-pytorch/torchft;

· jinsun-yoo opened on Jun 11, 2026

Slow TorchFT checkpoint recovery

Status: Open.

#329 In meta-pytorch/torchft;

· PGG106 opened on Jun 7, 2026

PGTransport.__init__(timeout=...) is effectively ignored by checkpoint send/recv

Status: Open.

#323 In meta-pytorch/torchft;

· wanna-01 opened on May 6, 2026

SIGSEV during long running training when using asynchronous quorum

Status: Open.

#316 In meta-pytorch/torchft;

· yurigorokhov opened on Mar 12, 2026

[New Feature] Integrate Optional Optimizer State Aggregation into a Unified Optimizer Class

Status: Open.

#312 In meta-pytorch/torchft;

· Iacob-Alexandru-Andrei opened on Feb 3, 2026

torchft should support transferring statefuldataloader state from other workers

Status: Open.

#309 In meta-pytorch/torchft;

· TimothySeah opened on Jan 13, 2026

Configure expected_replicas to avoid running tasks with unexpected replicas.

Status: Open.

#302 In meta-pytorch/torchft;

· zhengchenyu opened on Dec 19, 2025

Example train_ddp.py breaks

Status: Open.

#295 In meta-pytorch/torchft;

· kasakun opened on Dec 2, 2025

ProcessGroupBabyNCCL - EOFError

Status: Open.

#284 In meta-pytorch/torchft;

· btian opened on Oct 16, 2025

Using get_optimizer_state_dict inside state_dict causes TorchFT to get stuck

Status: Open.

#282 In meta-pytorch/torchft;

· btian opened on Oct 15, 2025

DDP models are different when training is interrupted

Status: Open.

#276 In meta-pytorch/torchft;

· btian opened on Sep 29, 2025

[RFC] [Intel GPU] Extending TorchFT to Support Intel GPU with XCCL Backend

Status: Open.

#257 In meta-pytorch/torchft;

· siju-samuel opened on Aug 18, 2025