Search results

The test dependency torchpippy is merged into PyTorch and the repository is archived

Status: Open.

#3846 In huggingface/accelerate;

· yurivict opened on Nov 23, 2025

FSDP2 with lora take more memory than FSDP

Status: Open.

#3844 In huggingface/accelerate;

· Zhikaiiii opened on Nov 21, 2025

fix: allow for flattening and unflattening optimizer state dict using flatten_optimizer_state_dict

Status: Open.

#3839 In huggingface/accelerate;

· kmehant opened on Nov 12, 2025

GPT-OSS fails to load with FSPD2

Status: Open.

#3838 In huggingface/accelerate;

· qgallouedec opened on Nov 11, 2025

notebook_launcher Kaggle RuntimeError: CUDA error: initialization error

Status: Open.

#3833 In huggingface/accelerate;

· dinhngoc267 opened on Nov 6, 2025

Torchao fp8 fails if using accelerate config file with Trainer

Status: Open.

#3830 In huggingface/accelerate;

· shimizust opened on Nov 3, 2025

Does Accelerate automatically set the DataLoader’s sampler to a DistributedSampler?

Status: Open.

#3829 In huggingface/accelerate;

· caixxiong opened on Nov 3, 2025

Accelerate mixed torch.Tensor and DTensor error when using TE FP8 and FSDP/TP

Status: Open.

#3827 In huggingface/accelerate;

· gmintoco opened on Nov 1, 2025

Accelerate save_state() error using FSDP2/TP

Status: Open.

#3826 In huggingface/accelerate;

· gmintoco opened on Nov 1, 2025

Error generating example: 'weight' must be 2-D in model.generate()

Status: Open.

#3825 In huggingface/accelerate;

· yurkoff-mv opened on Nov 1, 2025

load_checkpoint_and_dispatch not work

Status: Open.

#3824 In huggingface/accelerate;

· ziliangok opened on Oct 31, 2025

<code>TypeError: NoneType object is not callable</code> occurs when using <code>fsdp_auto_wrap_policy=NO_WRAP</code> with <code>fsdp_activation_checkpointing=true</code>

Status: Open.

#3822 In huggingface/accelerate;

· sdjasj opened on Oct 30, 2025