costinstroie/lfm-medical-1.2b

lfm-medical-1.2b (GGUF)

A medical-domain LoRA fine-tune merged onto [LiquidAI/LFM2.5-1.2B-Thinking](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Thinking), quantized to GGUF for use with llama.cpp.

Provenance

• —Base model: LiquidAI/LFM2.5-1.2B-Thinking
• —LoRA adapter: burtenshaw/unsloth-lfm-medical-lr-5e-5-20260211b (originally trained against unsloth/LFM2.5-1.2B-Instruct; re-merged here onto the Thinking checkpoint since it shares the same Lfm2ForCausalLM architecture/tokenizer)
• —Merge: peft merge_and_unload(), bfloat16
• —Conversion/quantization: llama.cpp's convert_hf_to_gguf.py (F16) → llama-quantize

Files

Usage

llama-cli -m lfm-medical-1.2b-Q4_K_M.gguf -p "What is C. elegans?"

Since the base is the Thinking variant, output includes <think>...</think> reasoning traces before the final answer.

Disclaimer

This is a research/experimental fine-tune. Not validated for clinical use.