Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

UNet2DModel dtype property fails under nn.DataParallel with UnboundLocalError in get_parameter_dtype

Open Beginner friendly
  • Is something wrong?
#13,789 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
74/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Quiet
Tech stack
python, pytorch
Domain
machine-learning

Research direction

Start in diffusers/models/modeling_utils.py at get_parameter_dtype and its find_tensor_attributes annotation, then follow the self.dtype access from diffusers/models/unets/unet_2d.py. Reproduce the failure with the provided UNet2DModel and torch.nn.DataParallel example, and add or run a regression check showing the forward pass completes without UnboundLocalError.

Written by the indexing model from the issue text.

Description

bug models
Describe the bug

Description

UNet2DModel fails under torch.nn.DataParallel during the forward pass when it accesses self.dtype inside UNet2DModel.forward.

The error appears to come from diffusers.models.modeling_utils.get_parameter_dtype, where the nested function annotation uses tuple:

def find_tensor_attributes(module: nn.Module) -> list[tuple[str, Tensor]]:

### Reproduction

import torch
from diffusers import UNet2DModel

print("torch:", torch.__version__)

import diffusers
print("diffusers:", diffusers.__version__)

model = UNet2DModel(
    sample_size=128,
    in_channels=10,
    out_channels=5,
    layers_per_block=2,
    block_out_channels=(64, 128, 256, 256),
    down_block_types=(
        "DownBlock2D",
        "DownBlock2D",
        "AttnDownBlock2D",
        "DownBlock2D",
    ),
    up_block_types=(
        "UpBlock2D",
        "AttnUpBlock2D",
        "UpBlock2D",
        "UpBlock2D",
    ),
    norm_num_groups=8,
)

model = model.cuda()
model = torch.nn.DataParallel(model, device_ids=[0, 1])
model.eval()

x = torch.randn(8, 10, 128, 128, device="cuda")
t = torch.rand(8, device="cuda") * 1000

with torch.no_grad():
    y = model(x, t, return_dict=False)[0]

print(y.shape)

### Logs

```shell
File ".../diffusers/models/unets/unet_2d.py", line ..., in forward
    t_emb = t_emb.to(dtype=self.dtype)

File ".../diffusers/models/modeling_utils.py", line ..., in dtype
    return get_parameter_dtype(self)

File ".../diffusers/models/modeling_utils.py", line ..., in get_parameter_dtype
    def find_tensor_attributes(module: nn.Module) -> list[tuple[str, Tensor]]:

UnboundLocalError: cannot access local variable 'tuple' where it is not associated with a value
System Info

Python 3.11.13, torch 2.9.1+cu128, diffusers 0.38.0, CUDA Version: 12.9

Who can help?

No response

Dominant language
Python
Stars
34.6k
Forks
7.4k
Avg merge
4d 17h
Merged PRs (30d)
48

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from huggingface/diffusers

All issues in huggingface/diffusers

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.