Llama-3-8B Message Rewriter
Custom fine-tuned Llama-3-8B for rewriting daily communication messages.
Clean 16-bit weights, GGUF-compatible.
Training
- Base Model: meta-llama/Meta-Llama-3-8B-Instruct
- Method: QLoRA, then dequantized to pure FP16
- Hardware: Free Kaggle T4 GPU
