shri171981/medical_chat_generative

HACK_DOC (Medical Assistant)

This is a Llama-3 based medical assistant model. It was fine-tuned on the ChatDoctor-HealthCareMagic-100k dataset to provide empathetic, doctor-style responses to medical queries.

Model Details

• Base Model: unsloth/llama-3-8b-instruct-bnb-4bit
• Adapter Type: LoRA (Rank 64)
• Training Framework: Unsloth / PyTorch
• Trigger Phrase: The model is trained to start responses with: "THANKS FOR ASKING HACK_DOC. Here is my answer:"

How to use

You must load this adapter on top of the base Llama-3 model.

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

# 1. Load Base Model
base_model_name = "unsloth/llama-3-8b-instruct-bnb-4bit"
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    load_in_4bit=True,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(base_model_name)

# 2. Load HACK_DOC Adapter
model = PeftModel.from_pretrained(base_model, "shri171981/genai_hack_doc")

# 3. Run Inference
inputs = tokenizer("I have a severe headache.", return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=128)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))