llama2-sdtm-ae-qlora

LLaMA-2 SDTM AE QLoRA Adapter

This repository contains a QLoRA adapter fine-tuned for:

Raw AE → SDTM AE mapping with Chain-of-Thought reasoning

Base Model

• meta-llama/Llama-2-7b-hf (gated)

Domain

• CDISC SDTM
• AE (Adverse Events)
• All therapeutic areas

Training

• QLoRA (4-bit NF4)
• LoRA rank: 64
• BF16 (A100)
• CSV-based CoT dataset

Usage

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

base = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

model = PeftModel.from_pretrained(
    base,
    "karamalanagendra/llama2-sdtm-ae-qlora"
)

tokenizer = AutoTokenizer.from_pretrained(
    "meta-llama/Llama-2-7b-hf"
)

prompt = "Map adverse event start date to SDTM"

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
out = model.generate(**inputs, max_new_tokens=200)

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