Skip to content

README.md

Latest commit

 

History

History
43 lines (29 loc) · 909 Bytes

README.md

File metadata and controls

43 lines (29 loc) · 909 Bytes
Raw

TechNavi LoRA Fine-tuning

Fine-tune GPT-OSS-20B with TechNavi YouTube channel knowledge using LoRA on Apple Silicon.

Quick Start

Run the complete demo:

./run_full_demo.sh

Or run steps individually:

# 1. Setup
./01_setup_environment.sh && source venv/bin/activate

# 2. Install MLX
./02_install_mlx.sh

# 3. Train LoRA
./04_run_lora_training.sh

# 4. Test model
python inference.py --adapter-path ./gpt_oss_lora_adapter --prompt "What is TechNavi?"

# 5. Compare base vs LoRA
./08_compare_lora_inference.sh

What it does

  • Before: Generic model with no TechNavi knowledge
  • After: Model knows TechNavi is a YouTube channel with reasoning capabilities

Key Features

  • ✅ Performance statistics (tokens/sec, memory usage)
  • ✅ Base vs LoRA comparison
  • ✅ Interactive mode

Training Data

52 examples about TechNavi YouTube channel with reasoning patterns.