Skip to content

MODELS.md

Latest commit

 

History

History
31 lines (21 loc) · 798 Bytes

MODELS.md

File metadata and controls

31 lines (21 loc) · 798 Bytes
Raw

Model Information

Pre-trained Model

The LoRA adapter files are excluded from this repository due to size (35MB).

To use this project:

  1. Run the training script to generate your own adapter:

    ./04_run_lora_training.sh

  2. Or download pre-trained adapters from:

    • [Add your model hosting URL here]

Model Details

  • Base Model: lmstudio-community/gpt-oss-20b-MLX-8bit
  • LoRA Adapter Size: ~35MB
  • Training Data: 52 TechNavi YouTube channel examples
  • Training Time: ~15 minutes on Apple Silicon
  • Performance: 87% loss reduction

Files Generated

After training, you'll have:

  • gpt_oss_lora_adapter/ - The fine-tuned LoRA weights
  • Various checkpoint files during training

These files are automatically excluded from git via .gitignore.