LoRA adapters fine-tuned on medical instruction data for the HealthMate-AI project (a medical RAG chatbot). The base model is mistralai/Mistral-7B-Instruct-v0.2; only adapter weights are stored in this repo.

Three adapters trained with identical hyperparameters but different random seeds are included, for robustness/variance analysis:

Adapter	Location
Seed 42 (default)	repo root
Seed 123	`adapter_seed_123/`
Seed 999	`adapter_seed_999/`

LoRA configuration

Rank (r): 16, alpha: 32, dropout: 0.05
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Task type: CAUSAL_LM, epochs: 3

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained(
    "mistralai/Mistral-7B-Instruct-v0.2", device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("yogvidwankhede/healthmate-mistral-7b-medical-lora")

# Seed 42 adapter (repo root)
model = PeftModel.from_pretrained(base, "yogvidwankhede/healthmate-mistral-7b-medical-lora")

# Or a specific seed:
# model = PeftModel.from_pretrained(base, "yogvidwankhede/healthmate-mistral-7b-medical-lora", subfolder="adapter_seed_123")

Intended use and limitations

Research/educational use as part of a retrieval-augmented medical assistant. Not a substitute for professional medical advice, diagnosis, or treatment. Outputs may be inaccurate; always verify with qualified healthcare professionals.