Add custom dataset fine-tuning workflow for Cosmos Reason 1 #33
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This contribution adds a complete workflow for fine-tuning Cosmos Reason 1 on custom datasets with local video files and human-labeled quality scores. Added components: - Dataset preparation scripts for creating train/eval splits - Conversation format conversion for SFT training - Model evaluation script with HTML report generation - Training configuration for custom datasets - Documentation in physical plausibility recipe The workflow supports: - Stratified train/eval splitting with label balancing - Label scaling and distribution management - Integration with Cosmos Transfer-generated videos - Comprehensive evaluation metrics and reporting This extends the existing VideoPhy-2 recipe to enable practitioners to fine-tune on domain-specific video quality assessment tasks. Signed-off-by: Saurav Nanda <sauravn@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This PR adds a complete workflow for fine-tuning Cosmos Reason 1 on custom datasets with local video files and human-labeled physical plausibility scores.
Added Components
create_dataset_with_split.py,add_conversations_to_dataset.py)evaluate_model.py)custom_dataset_sft_config.toml)Features
Context
Extends the existing VideoPhy-2 recipe in the physical plausibility post-training guide to enable practitioners to fine-tune on domain-specific video quality assessment tasks. The workflow follows cookbook conventions where users copy scripts to their cosmos-reason1 workspace.
Files Changed
docs/recipes/post_training/reason1/physical-plausibility-check/post_training.md- Added custom dataset sectiondocs/recipes/post_training/reason1/physical-plausibility-check/assets/custom_dataset_sft_config.toml- New training configscripts/examples/reason1/physical-plausibility-check/create_dataset_with_split.py- New dataset prep scriptscripts/examples/reason1/physical-plausibility-check/add_conversations_to_dataset.py- New format converterscripts/examples/reason1/physical-plausibility-check/evaluate_model.py- New evaluation script