torchrun --nproc_per_node=8 generate.py wan2.2_i2v --parallel ulysses_tp --ulysses-anything --num-frames 81 --model-path /sse_ard/pretrained_models/Wan2.2-I2V-A14B-Diffusers --cache --num_inference_steps 2 --track-memory
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.3k
- Forks
- 88
- Avg merge
- 1d 4h
- Merged PRs (30d)
- 14
Description
[rank3]: File "/usr/local/lib/python3.10/dist-packages/diffusers/models/_modeling_parallel.py", line 123, in setup
[rank3]: self._ring_mesh = self._mesh["ring"]
[rank3]: File "/usr/local/lib/python3.10/dist-packages/torch/distributed/device_mesh.py", line 712, in getitem
[rank3]: slice_mesh_dims = _mesh_resources._get_slice_mesh_dims(
[rank3]: File "/usr/local/lib/python3.10/dist-packages/torch/distributed/device_mesh.py", line 295, in _get_slice_mesh_dims
[rank3]: raise RuntimeError("Cannot create a submesh from a submesh.")
[rank3]: RuntimeError: Cannot create a submesh from a submesh.
usp + tp好像不支持？
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Reproduce the command in generate.py with --parallel ulysses_tp and --ulysses-anything, then inspect the traceback around diffusers/models/_modeling_parallel.py:123 and PyTorch device_mesh.py. Determine whether USP and tensor parallelism are supported together; done should be a confirmed diagnosis and a project-level resolution or documented limitation.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- distributed-systems, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100