LLM Fine-Tuning & Local Inference
Engineering work by Orçun Candan.
Implementation
Qwen2.5-Coder 0.5B logo model with LoRA experiments, a custom PyTorch training loop, GGUF quantization and CPU inference.
Results & scope
Recorded CPU benchmark: 24/24 valid outputs, 21/23 brand names preserved, 26.2 s mean generation time. Small synthetic-data experiment.
Tools & technologies
- Qwen2.5-Coder
- PyTorch
- LoRA
- GGUF
- CPU inference