🧬 LoRA Fine-Tune Lab

Parameter-efficient fine-tuning of Mistral-7B using LoRA, demonstrating production-grade ML engineering: dataset preparation, 4-bit quantized training, evaluation with automated metrics, and efficient inference serving.

---

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        LoRA Fine-Tune Lab                       │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────────┐  │
│  │  Data Prep   │───▶│  Fine-Tune   │───▶│    Evaluation    │  │
│  │              │    │              │    │                  │  │
│  │ • Load HF    │    │ • QLoRA 4bit │    │ • Perplexity     │  │
│  │   dataset    │    │ • r=16 α=32  │    │ • ROUGE scores   │  │
│  │ • Tokenize   │    │ • Grad accum │    │ • Task metrics   │  │
│  │ • Train/val  │    │ • Checkpoints│    │ • W&B logging    │  │
│  │   split      │    │              │    │                  │  │
│  └──────────────┘    └──────┬───────┘    └──────────────────┘  │
│                             │                                   │
│                    ┌────────▼────────┐                          │
│                    │   LoRA Adapter  │                          │
│                    │   (~16 MB)      │                          │
│                    └────────┬────────┘                          │
│                             │                                   │
│                    ┌────────▼────────┐                          │
│                    │   Inference     │                          │
│                    │   CLI / API     │                          │
│                    │                 │                          │
│                    │ Base + Adapter  │                          │
│                    │ 4-bit quantized │                          │
│                    └─────────────────┘                          │
│                                                                 │
├─────────────────────────────────────────────────────────────────┤
│  Tech: Transformers · PEFT · Accelerate · bitsandbytes · W&B   │
└─────────────────────────────────────────────────────────────────┘

What This Demonstrates

Skill	Implementation
Parameter-efficient fine-tuning	LoRA with rank-16 decomposition on attention projections
Memory-efficient training	QLoRA 4-bit quantization via bitsandbytes NF4
Training engineering	Gradient accumulation, cosine scheduling, warmup
Evaluation rigor	Perplexity, ROUGE-L, task-specific accuracy
Experiment tracking	Weights & Biases integration with artifact logging
Production inference	Merged adapter loading with CLI interface
Reproducibility	YAML configs, fixed seeds, pinned dependencies

Quick Start

# Clone and install
git clone https://github.com/YOUR_USER/lora-finetune-lab.git
cd lora-finetune-lab
pip install -r requirements.txt

# Prepare data
python src/data_prep.py --dataset tatsu-lab/alpaca --output data/

# Fine-tune (requires GPU — see GUIDE.md)
python -m notebooks.finetune

# Evaluate
python -m notebooks.evaluate

# Run inference
python src/inference.py --adapter outputs/lora-adapter --prompt "Explain LoRA in one paragraph."

Project Structure

lora-finetune-lab/
├── configs/
│   ├── lora_config.yaml          # LoRA hyperparameters
│   └── training_config.yaml      # Training loop config
├── notebooks/
│   ├── finetune.py               # Fine-tuning notebook (percent script)
│   └── evaluate.py               # Evaluation notebook (percent script)
├── src/
│   ├── data_prep.py              # Dataset preprocessing
│   └── inference.py              # Inference CLI
├── requirements.txt
├── GUIDE.md                      # Step-by-step execution guide
└── README.md

Results

Run Date: February 12, 2026 Hardware: NVIDIA A100 Training Time: ~3 hours (2h 59m 54s) WandB Report: View Run Logs

Metric	Value	Notes
Eval Loss	1.187	Low validation loss indicates good generalization
Eval Entropy	0.888	Measure of prediction uncertainty
Token Accuracy	70.78%	Percentage of tokens predicted correctly
Train Loss	0.890	Final training loss after 3 epochs

Example Inference:

Instruction: Write a python function to reverse a list.

Response:

def reverse_list(lst):
   return lst[::-1]

Training Details:

• Base model: mistralai/Mistral-7B-v0.3
• Trainable parameters: ~4.2M / 7.2B (0.06%)
• Epochs: 3
• Speed: 13.73 samples/sec (train), 61.35 samples/sec (eval)

License

MIT