| --- |
| tags: |
| - gguf |
| - llama.cpp |
| - unsloth |
| --- |
| |
| # ZPT-Commit-1.2b-Instruct : Git Commit Message Generator |
|
|
| ## 🚀 Overview |
|
|
| **ZPT-Commit-1.2B-Instruct** is a highly specialized, 1.2 Billion parameter language model engineered specifically to automate and improve the creation of descriptive and professional Git commit messages. |
|
|
| Fine-tuned from the robust **LiquidAI LFM 2.5 1.2b** base model, this model uses the `Tavernari/git-commit-message-dt` dataset to excel at analyzing code diffs and transforming technical changes into clear, actionable commit summaries. It is designed to streamline development workflows by ensuring a clean, searchable, and informative git history. |
|
|
| --- |
|
|
| ## 🛠️ Technical Specifications |
|
|
| This model is provided in highly optimized formats for maximum efficiency across various hardware. |
|
|
| | Feature | Specification | Details | |
| | :--- | :--- | :--- | |
| | **Model Name** | ZPT-Commit-1.2B-Instruct | Specialized version for generating precise commit messages. | |
| | **Base Model** | LiquidAI LFM 2.5 1.2b | The foundational architecture. | |
| | **Training Data** | `Tavernari/git-commit-message-dt` | Dataset used to fine-tune the model on real-world code changes and commit patterns. | |
| | **Fine-tuning Framework** | Unsloth Studio | Trained 2x faster using the Unsloth optimization techniques. | |
| | **Model Size** | 1.2 Billion Parameters | Offers strong performance while maintaining a manageable footprint for deployment. | |
| | **Supported Formats** | GGUF, Quantized | Optimized for CPU/GPU inference via `llama.cpp`. | |
|
|
| --- |
|
|
| ## ⚙️ Performance & Implementation |
|
|
| The model has been converted and optimized to the **GGUF** format, allowing for highly efficient local deployment without needing massive GPU resources. |
|
|
| **Performance Benefits:** |
| * **Unsloth Optimization:** The training process utilized Unsloth, enabling faster training cycles and an efficient model structure. |
| * **GGUF Efficiency:** Quantized files ensure excellent inference speed and lower memory usage, making it ideal for CI/CD pipelines or local development environments. |
|
|
| **Available Model Files:** |
| The following optimized file is available for immediate use: |
| * `LFM2.5-1.2B-Instruct.Q8_0.gguf` |
|
|
| --- |
|
|
| ## 💡 Use Cases |
|
|
| ZPT-Commit-1.2B-Instruct is designed to be a powerful aid in the software development lifecycle. |
|
|
| * **Automated Commit History:** Automatically generating high-quality, structured commit messages directly from code diffs. |
| * **Developer Workflow Integration:** Integrating into IDE extensions or CI/CD tools to enforce consistent commit message standards. |
| * **Code Review Enhancement:** Providing context-rich summaries of changes, speeding up the review process. |
| * **Clean Repository Management:** Maintaining a professional and easily traceable version control history. |
|
|
| --- |
|
|
| ## 💻 Installation & Usage |
|
|
| The model is designed to work seamlessly with the `llama.cpp` ecosystem. |
|
|
| **Dependencies:** |
| * `llama.cpp` (for CPU/GPU inference) |
| * `unsloth` (for initial conversion and optimization) |
|
|
| **CLI Usage:** |
| You can invoke the model using the following command structure. The model will require the code diff as input for the best results. |
|
|
| * **Text-Only LLMs:** |
| ```bash |
| llama-cli -hf zmail-tech/ZPT-Commit-1.2B-Instruct --jinja |
| ``` |
| * **Multimodal Models:** |
| ```bash |
| llama-mtmd-cli -hf zmail-tech/ZPT-Commit-1.2B-Instruct --jinja |
| ``` |
| |
| --- |
| ***Note:** The model is named **Commit** as its primary and specialized function is the generation of Git commit messages. For full setup instructions, please refer to the [Unsloth AI GitHub](https://github.com/unslothai/unsloth) resources.* |
