You can deploy this model and start earning money today!

yammysquid/
medical_Llama_3.1_8B

Name: medical_Llama_3.1_8B
Availability: InStock

safetensors

meta-llama/Llama-3.1-8B-Instruct

finetune

Views

No views yet

This repository hosts a LoRA adapter that fine-tunes meta-llama/Meta-Llama-3.1-8B-Instruct for clinical question answering. The adapter encourages explicit chain-of-thought reasoning so the model walks through diagnostic steps before giving a final answer.

Model Details

Base model: Meta-Llama-3.1-8B-Instruct
Adapter type: PEFT LoRA (r=16, alpha=16, dropout=0.0)
Precision: Base weights loaded in 8-bit via BitsAndBytes during training
Training data: FreedomIntelligence/medical-o1-reasoning-SFT (English split with Question, Complex_CoT, Response fields)
Context window used in training: 2,048 tokens
Evaluation: Manual spot-checks on representative clinical prompts (no automated metrics reported)

Intended Use & Limitations

Designed for educational or research explorations of medical reasoning with large language models.
Not suitable for real clinical decision making without expert oversight.
Outputs may include hallucinated facts, outdated guidelines, or unsafe treatment suggestions. Always consult licensed medical professionals before acting on any response.

Prompt Format

Training used the following template:

Please provide an appropriate response based on the instruction and the specific question given below.
Before answering, please think carefully about the question and demonstrate a step-by-step reasoning process to ensure the logic and accuracy of your response.

### Instruction:
You are a medical expert proficient in clinical reasoning, diagnosis, and treatment planning.
Please answer the medical question below.

### Question:
{Question}

### Answer:
<think>
{Complex_CoT}
</think>
{Response}

You can follow the same structure or adapt it to your use case. The <think> tags are optional but help separate the reasoning trace from the final answer.

How to Load the Adapter

python

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
from peft import PeftModel

base_model = "meta-llama/Meta-Llama-3.1-8B-Instruct"
adapter_id = "your-username/medical-llm-reasoning-lora"  # replace with this repo ID

quant_cfg = BitsAndBytesConfig(load_in_8bit=True)
model = AutoModelForCausalLM.from_pretrained(
    base_model,
    quantization_config=quant_cfg,
    device_map="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
if tokenizer.pad_token is None:
    tokenizer.pad_token = tokenizer.eos_token

model = PeftModel.from_pretrained(model, adapter_id)

prompt = """Please provide an appropriate response based on the instruction and the question.
Instruction: You are a medical expert proficient in clinical reasoning.
Question: A patient has a persistent cough, hemoptysis, low-grade fever, and night sweats. What is the most likely diagnosis?
Answer:
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    eos_token_id=tokenizer.eos_token_id,
    temperature=0.7,
    top_p=0.9,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

yammysquid/medical_Llama_3.1_8B

Model Details

Intended Use & Limitations

Prompt Format

How to Load the Adapter

yammysquid/
medical_Llama_3.1_8B