Skip to content

Add HuggingFaceCheckpointer option for only registering final checkpoint #1516

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 16 commits into from
Sep 12, 2024

Conversation

irenedea
Copy link
Contributor

@irenedea irenedea commented Sep 10, 2024

This would allow us to avoid saving an extra final checkpoint at the end of training.

Example Usage

  hf_checkpointer:
    precision: bfloat16
    save_folder: ./local_checkpoints
    save_interval: 1h
    mlflow_logging_config:
      task: llm/v1/chat
      metadata:
        task: llm/v1/chat
        pretrained_model_name: meta-llama/Meta-Llama-3-8B
        databricks_model_family: LlamaForCausalLM (llama-3)
        databricks_model_source: genai-fine-tuning
        databricks_model_size_parameters: 8b
    mlflow_registered_model_name: ift-meta-llama-3-8b-ohuja8
    final_register_only: true

Branch for testing: final-register-only-fallback This adds irenedea@f63b5e0 which raises an error.

fallback: test-final-register-only-fallback-ATwQG5
normal: test-final-register-only-SKerij

@irenedea irenedea requested a review from a team as a code owner September 10, 2024 19:54
@irenedea irenedea marked this pull request as draft September 10, 2024 19:54
@irenedea irenedea marked this pull request as ready for review September 11, 2024 00:35
@irenedea irenedea requested a review from KuuCi September 11, 2024 00:36
Copy link
Contributor

@KuuCi KuuCi left a comment

Choose a reason for hiding this comment

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

lgtm aside from that one snipbit I don't quite understand.

Copy link
Contributor

@snarayan21 snarayan21 left a comment

Choose a reason for hiding this comment

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

seems chill to me, would like @dakinggg to take a pass

Copy link
Collaborator

@dakinggg dakinggg left a comment

Choose a reason for hiding this comment

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

one typo i think, otherwise lgtm

@irenedea irenedea enabled auto-merge (squash) September 12, 2024 23:32
@irenedea irenedea merged commit a862d6e into mosaicml:main Sep 12, 2024
9 checks passed
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.

4 participants