Skip to content
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

Cli . py #1035

Open
wants to merge 7 commits into
base: main
Choose a base branch
from
Open

Cli . py #1035

wants to merge 7 commits into from

Conversation

Leoleojames1
Copy link

The unsloth-cli.py is a powerful interface for building local applications with Unsloth. After building the CLI into the backend of my dataset construction tool, I realized that heavy modifications were needed in order to support the finetuning pipeline I had constructed. The following is my proposal for unsloth-cli-2.py, an enhanced feature set for the unsloth CLI:

Key Features:

  1. Flexible Data Handling:

    • Support for Parquet, JSON, and other data formats
    • Custom data parsing and processing pipelines
  2. Advanced Model Management:

    • Load and save models in various formats (Hugging Face, GGUF)
    • Quantization options for memory-efficient training and inference
    • Dequantization capabilities for precision-sensitive operations
  3. Comprehensive Training Pipeline:

    • Support for train, validation, and test datasets
    • Customizable training parameters (batch size, learning rate, etc.)
    • Integration with popular optimization techniques (LoRA, gradient checkpointing)
  4. Merging and Adaptation:

    • Merge LoRA adapters with base models
    • Dequantization options for merging quantized models
  5. Deployment and Sharing:

    • GGUF conversion for optimized model deployment
    • Direct integration with Hugging Face Hub for easy model sharing
  6. Robust Error Handling and Logging:

    • Detailed error messages and logging for easier debugging
    • Fallback options for model loading and processing

unsloth-cli.py enhanced feature set, via unsloth-cli-2.py, the scope of the feature set includes:
- Comprehensive training pipeline with validation and testing support
- Advanced error handling and fallback options for robust model loading
- Merging functionality for LoRA adapters with dequantization options
- Flexible quantization and precision control (4-bit, 16-bit, 32-bit)
- Support for custom datasets and data formats (JSON, Parquet)
- Integration with Hugging Face models and push-to-hub functionality
- GGUF conversion for optimized model deployment
- Enhanced logging and progress tracking
tabs/line spacing issue
★🚀commit message, looks nice🚀★
@shimmyshimmer shimmyshimmer changed the title [email protected] Cli . py Oct 3, 2024
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.

1 participant