Fine-tuned small language model adapter (meta-llama/Llama-3.2-3B-Instruct or microsoft/Phi-3.5-mini-instruct) trained on Spanish Patient Leaflets (Prospectos) and Summaries of Product Characteristics (Fichas Técnicas) from the AEMPS CIMA Research Dataset (hsilvosa/aemps-cima) and (hsilvosa/openplacsp).

Acts as a grounded Spanish medical QA assistant for active ingredients, administration routes, side effects, contraindications, and excipient safety warnings.

Usage

python

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

base_model = "meta-llama/Llama-3.2-3B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype="auto", device_map="auto")
model = PeftModel.from_pretrained(model, "your-hf-username/CIMA-Spanish-Medical-Llama-LoRA")

prompt = "<|system|>
Eres un asistente médico farmacéutico...
<|user|>
¿Cuál es la vía de administración y composición de Omeprazol?
<|assistant|>
"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=150)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

hsilvosa/CIMA-Spanish-Medical-Llama-LoRA

CIMA Spanish Medical Llama (LoRA)

Usage

hsilvosa/
CIMA-Spanish-Medical-Llama-LoRA