Model Card for SmolLM2-360M-MedReason

Model Details

This model is a fine-tuned version of SmolLM2-360M-Instruct on a medical reasoning dataset; UCSC-VLAA/MedReason

Model Description

• Developed by: Rustam Shiriyev
• Model type: Instruction-tuned model
• Language(s) (NLP): English
• License: MIT
• Finetuned from model: unsloth/SmolLM2-360M-Instruct

How to Get Started with the Model

from peft import PeftModel
from huggingface_hub import login
from transformers import AutoTokenizer

login(token="")

tokenizer = AutoTokenizer.from_pretrained("unsloth/SmolLM2-360M-MedReason",)

base_model = AutoModelForCausalLM.from_pretrained(
    "unsloth/SmolLM2-360M-MedReason",
    device_map={"": 0}, token=""
)

model = PeftModel.from_pretrained(base_model,"Rustamshry/SmolLM2-360M-MedReason")

question = "Which of the following nipple discharge is most probably physiological?"
options = """Answer Choices:
A. B/L spontaneous discharge
B. B/L milky discharge with squeezing from multiple ducts
C. U/L bloody discharge
D. U/L bloody discharge with squeezing from a single duct"""

prompt = f"""### Question:\n{question}\n{options}\n\n### Response:\n"""

input_ids = tokenizer(prompt, return_tensors="pt").to(model.device)
   outputs = model.generate(
   **input_ids,
   max_new_tokens=2000,
   #temperature=0.6,
   #top_p=0.95,
   #do_sample=True,
   #eos_token_id=tokenizer.eos_token_id
)

print(tokenizer.decode(outputs[0]),skip_special_tokens=True)

Framework versions

• PEFT 0.14.0