You can deploy this model and start earning money today!

SimonLeclere/
Meta-Llama-3.1-8B-Instruct-Diff2Commit-Q4_K_M

Name: Meta-Llama-3.1-8B-Instruct-Diff2Commit-Q4_K_M
Availability: InStock

gguf

endpoints_compatible

conversational

Views

No views yet

Model Card: Llama 3.1 8B Instruct - Git Commit Message Generator

Model Details

Model Description

This is a fine-tuned version of Meta-Llama-3.1-8B-Instruct optimized for generating concise, human-readable Git commit messages from code diffs. The model has been trained using Quantized Low-Rank Adaptation (QLoRA) and quantized to 4-bit precision using Unsloth for efficient local deployment.

Model Family: Llama 3.1 8B Instruct
Fine-tuning Method: QLoRA (Quantized Low-Rank Adaptation)
Quantization: 4-bit (Q4_K_M) via Unsloth
Model Size: 4.92 GB (quantized)
Training Framework: Unsloth + TRL + Hugging Face Transformers

Model Sources

Base Model: unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit
Quantization: Unsloth

Training Data

The model was trained on a diverse dataset combining:

Diffs from real commits from popular open-source Python repositories:
- Django, Flask, Requests, Pandas, NumPy, scikit-learn, PyTorch, TensorFlow/Keras, matplotlib, pytest, FastAPI, Transformers, and others
Existing dataset: seniruk/git-diff_to_commit_msg from Hugging Face
Synthetic data generated using larger LLMs to complete the dataset with high-quality commit messages.

License

This model inherits the license of its base model (Meta-Llama-3.1-8B-Instruct). Please refer to the Meta License Agreement for details.

SimonLeclere/Meta-Llama-3.1-8B-Instruct-Diff2Commit-Q4_K_M