juzhengz

/

LoRI-S_code_llama3_rank_64

Like 0

This PR significantly enhances the model card for tomg-group-umd/LoRI-S_code_llama3_rank_64 by adding comprehensive information and improving its discoverability and usability.

Key updates include:

• Metadata Enrichment: Adding license: apache-2.0, and relevant tags such as peft, lora, code-generation, and llama, along with the specific datasets used for training this model.
• Detailed Model Description: Populating the "Model Details" section with information about the developers, model type, language, and the base model, based on the paper abstract and GitHub repository.
• Complete Model Sources: Adding direct links to the official GitHub repository, the Hugging Face paper page, the project page, and the Hugging Face collection.
• Elaborated Usage Instructions: Filling in "Uses" sections (Direct Use, Downstream Use, Out-of-Scope) to clarify the model's intended applications and limitations.
• Executable Code Snippet: Providing a runnable Python code example in "How to Get Started" for quick inference using transformers and peft.
• Training Information: Detailing the "Training Data" and "Training Procedure" (LoRI-D and LoRI-S stages, FSDP) and "Training Hyperparameters" (rank, sparsity, etc.).
• Evaluation Summary: Summarizing key evaluation aspects and directing users to the paper for detailed results.
• Citation: Including the BibTeX entry from the paper.
• Visual Aid: Embedding the LoRI architecture diagram from the GitHub repository.

This update makes the model card much more informative and user-friendly for researchers and practitioners.