LLaMA-3.1-8B Pharmaceutical Compliance QLoRA

Fine-tuned LLaMA-3.1-8B-Instruct on 500 FDA pharmaceutical compliance instruction pairs using QLoRA + Unsloth.

Model Details

Property Value
Base Model meta-llama/Meta-Llama-3.1-8B-Instruct
Method QLoRA 4-bit NF4 + LoRA rank 64
Framework Unsloth (2x faster, 60% less VRAM)
Domain FDA CFR Title 21, GMP, ICH guidelines
Training Samples 450
Epochs 3
GPU A100 40GB

Domain Coverage

  • 21 CFR Part 210/211 (cGMP)
  • 21 CFR Part 201 (Drug labeling)
  • ICH Q8/Q9/Q10 (Pharmaceutical quality)
  • OOS investigations, CAPA writing
  • Deviation reports, process validation
  • Bioequivalence requirements

Usage

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="nihanthnaidu007/llama-3.1-8b-pharma-compliance-qlora-v4",
    max_seq_length=1024,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)

Author

Nihanth Naidu Kalisetti — AI Engineer W&B Run: https://wandb.ai/nihanthnaidu-kalisetti-long-island-university/llama-pharma-qlora/runs/vt1w00g4

Downloads last month
7
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for nihanthnaidu007/llama-3.1-8b-pharma-compliance-qlora-v4

Adapter
(1974)
this model

Space using nihanthnaidu007/llama-3.1-8b-pharma-compliance-qlora-v4 1