Skip to content

Conversation

@jzhang38
Copy link
Collaborator

@jzhang38 jzhang38 commented Jun 7, 2025

  1. remove fastvideo_args.device and fastvideo_args.device_str. Use get_torch_device() instead.
  2. remove unused functions in weight_utils.py
  3. prioritize using maybe_init_distributed_environment_and_model_parallel
  4. move check_fastvideo_args inside post_init

@jzhang38 jzhang38 marked this pull request as draft June 7, 2025 20:52
@jzhang38 jzhang38 marked this pull request as ready for review June 7, 2025 21:35
@jzhang38 jzhang38 temporarily deployed to runpod-runners June 7, 2025 22:43 — with GitHub Actions Inactive
@jzhang38 jzhang38 temporarily deployed to runpod-runners June 7, 2025 22:43 — with GitHub Actions Inactive
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lets we just move this and related V1 preprocessing files (except those under scripts) to under fastvideo/v1?

fastvideo_args = FastVideoArgs(model_path=args.model_path,
num_gpus=world_size,
device_str="cuda",
num_gpus=get_world_size(),
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

perhaps we should just remove num_gpus altogether, including in inference? If we always hardcode num_gpus=get_world_size() users would need to control it using CUDA_VISIBLE_DEVICES. When I added num_gpus I was only considering inference and also assumed tp_size == sp_size

Copy link
Collaborator

@Edenzzzz Edenzzzz Jun 9, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think num_gpus is easier to control, if not specified we can use world size

@SolitaryThinker
Copy link
Collaborator

Thanks!

Copy link
Collaborator

@JerryZhou54 JerryZhou54 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@jzhang38 jzhang38 merged commit 46e7a15 into main Jun 8, 2025
7 checks passed
@jzhang38 jzhang38 deleted the py/improve_distributed_calling branch June 10, 2025 23:44
qimcis pushed a commit to qimcis/FastVideo that referenced this pull request Oct 30, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants