[Feature]Add VSA slurm training example scripts #666

BrianChen1129 · 2025-07-30T03:19:05Z

No description provided.

gemini-code-assist

Summary of Changes

Hello @BrianChen1129, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request introduces new Slurm training examples for Image-to-Video (I2V) and Text-to-Video (T2V) models that leverage Video Sparse Attention (VSA), enhancing the training capabilities within the project. It also includes a reorganization of existing distillation scripts and a utility for dataset acquisition, streamlining the setup process for users.

Highlights

New VSA Training Examples: Added dedicated Slurm scripts (I2V-14B-VSA.slurm, T2V-14B-VSA.slurm) for training Image-to-Video (I2V) and Text-to-Video (T2V) models using Video Sparse Attention (VSA). These scripts are configured with specific VSA-related arguments and environment setups for Slurm clusters.
Slurm Script Reorganization: Renamed and relocated existing distillation Slurm scripts (distill_dmd_VSA_t2v.slurm, distill_dmd_VSA_t2v_14B_480P.slurm, distill_dmd_t2v.slurm) to a new Wan2.1-T2V-480P directory, along with minor argument formatting adjustments within these files.
Dataset Download Utility: Included a new download_dataset.sh script to simplify the acquisition of 480P and 720P Wan-Syn datasets from Hugging Face, providing necessary data for the new training examples.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

This pull request adds several new example scripts for VSA (Video Sparse Attention) training using Slurm. It also refactors some existing scripts. The new scripts are well-structured but have some minor issues like redundant environment variable exports, non-portable home directory paths, and missing newlines at the end of files. There is also a critical typo in one of the renamed scripts that would break its execution. My review provides suggestions to fix these issues.

gemini-code-assist · 2025-07-30T03:21:39Z

examples/distill/Wan2.1-T2V-480P/distill_dmd_VSA_t2v_1.3B.slurm

 training_args=(
  --tracker_project_name wan_t2v_distill_dmd_VSA
-  --output_dir="checkpoints/wan_t2v_finetune"
+  --output_dir"checkpoints/wan_t2v_finetune"


There seems to be a typo here. The = was removed but not replaced with a space. This will likely cause the argument to be parsed incorrectly as --output_dircheckpoints/wan_t2v_finetune, which would break the script. Based on other changes in this PR, the intention was likely to separate the argument and its value with a space.

--output_dir "checkpoints/wan_t2v_finetune"

gemini-code-assist · 2025-07-30T03:21:39Z

examples/training/finetune/Wan2.1-VSA/I2V-14B-VSA.slurm

+set -e -x
+
+# Environment Setup
+source ~/conda/miniconda/bin/activate


Using ~ for the home directory might not be portable in all shell environments, especially non-interactive ones. It's generally safer and more explicit to use $HOME.

source $HOME/conda/miniconda/bin/activate

gemini-code-assist · 2025-07-30T03:21:39Z

examples/training/finetune/Wan2.1-VSA/I2V-14B-VSA.slurm

+export CUDA_VISIBLE_DEVICES=$SLURM_LOCALID
+export TOKENIZERS_PARALLELISM=false
+export WANDB_BASE_URL="https://api.wandb.ai"
+export WANDB_MODE=online


The WANDB_MODE environment variable is exported twice. This line is redundant as it's already set on line 20.

gemini-code-assist · 2025-07-30T03:21:39Z

examples/training/finetune/Wan2.1-VSA/I2V-14B-VSA.slurm

+    "${optimizer_args[@]}" \
+    "${validation_args[@]}" \
+    "${miscellaneous_args[@]}" \
+    "${vsa_args[@]}"  


It's a good practice to end files with a newline character for POSIX compatibility.

"${vsa_args[@]}"

gemini-code-assist · 2025-07-30T03:21:40Z

examples/training/finetune/Wan2.1-VSA/T2V-14B-VSA.slurm

+set -e -x
+
+# Environment Setup
+source ~/conda/miniconda/bin/activate


Using ~ for the home directory might not be portable in all shell environments, especially non-interactive ones. It's generally safer and more explicit to use $HOME.

source $HOME/conda/miniconda/bin/activate

gemini-code-assist · 2025-07-30T03:21:40Z

examples/training/finetune/Wan2.1-VSA/T2V-14B-VSA.slurm

+export CUDA_VISIBLE_DEVICES=$SLURM_LOCALID
+export TOKENIZERS_PARALLELISM=false
+export WANDB_BASE_URL="https://api.wandb.ai"
+export WANDB_MODE=online


The WANDB_MODE environment variable is exported twice. This line is redundant as it's already set on line 20.

gemini-code-assist · 2025-07-30T03:21:40Z

examples/training/finetune/Wan2.1-VSA/T2V-14B-VSA.slurm

+    "${optimizer_args[@]}" \
+    "${validation_args[@]}" \
+    "${miscellaneous_args[@]}" \
+    "${vsa_args[@]}"  


It's a good practice to end files with a newline character for POSIX compatibility.

"${vsa_args[@]}"

gemini-code-assist · 2025-07-30T03:21:40Z

examples/training/finetune/Wan2.1-VSA/download_dataset.sh

+python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x448x832_600k" --local_dir "FastVideo/Wan-Syn_77x448x832_600k" --repo_type "dataset"
+
+# 720P dataset
+python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x768x1280_250k" --local_dir "FastVideo/Wan-Syn_77x768x1280_250k" --repo_type "dataset"


It's a good practice to end files with a newline character for POSIX compatibility.

Suggested change

python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x768x1280_250k" --local_dir "FastVideo/Wan-Syn_77x768x1280_250k" --repo_type "dataset"

python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x768x1280_250k" --local_dir "FastVideo/Wan-Syn_77x768x1280_250k" --repo_type "dataset"

SolitaryThinker · 2025-07-30T05:23:59Z

examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P/distill_dmd_VSA_t2v_1.3B.slurm

 training_args=(
  --tracker_project_name wan_t2v_distill_dmd_VSA
-  --output_dir="checkpoints/wan_t2v_finetune"
+  --output_dir"checkpoints/wan_t2v_finetune"


space between arg name and arg?

BrianChen1129 added 20 commits June 30, 2025 04:10

update

bc8b24c

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

9c135a3

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

3973469

sy

fc63bec

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

e7c363d

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

9352197

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

e1b2c40

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

862ac8a

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

5c94cba

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

ebae132

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

6d446c6

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

df6afbb

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

a0522cd

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

708fe18

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

76006d2

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

3fc247d

Merge branch 'main' of github.com:hao-ai-lab/FastVideo

3efe6da

add VSA example

6ded688

update

ef2f2ed

update

e9416af

gemini-code-assist bot reviewed Jul 30, 2025

View reviewed changes

BrianChen1129 added 2 commits July 30, 2025 03:19

update

022fb72

update

ae7dc93

gemini-code-assist bot reviewed Jul 30, 2025

View reviewed changes

BrianChen1129 added the go Trigger Buildkite CI label Jul 30, 2025

BrianChen1129 added 4 commits July 30, 2025 03:38

update dataset example

d830f89

update

f78ecbf

update

627c03a

update

c764987

SolitaryThinker approved these changes Jul 30, 2025

View reviewed changes

BrianChen1129 merged commit f759980 into main Jul 30, 2025
1 check failed

BrianChen1129 deleted the yq/update branch August 12, 2025 22:23

qimcis pushed a commit to qimcis/FastVideo that referenced this pull request Oct 30, 2025

[Feature]Add VSA slurm training example scripts (hao-ai-lab#666)

569f8b7

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[Feature]Add VSA slurm training example scripts #666

[Feature]Add VSA slurm training example scripts #666

Uh oh!

BrianChen1129 commented Jul 30, 2025

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

gemini-code-assist bot Jul 30, 2025

Uh oh!

SolitaryThinker Jul 30, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x768x1280_250k" --local_dir "FastVideo/Wan-Syn_77x768x1280_250k" --repo_type "dataset"
	python scripts/huggingface/download_hf.py --repo_id "FastVideo/Wan-Syn_77x768x1280_250k" --local_dir "FastVideo/Wan-Syn_77x768x1280_250k" --repo_type "dataset"

[Feature]Add VSA slurm training example scripts #666

[Feature]Add VSA slurm training example scripts #666

Uh oh!

Conversation

BrianChen1129 commented Jul 30, 2025

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

SolitaryThinker Jul 30, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants