pranav-pvnn/codellama-7b-python-ai-assistant-LoRA-adapters

CodeLlama 7B Python AI Assistant (QLoRA)

Fine-tuned CodeLlama-7B model specialized for Python programming assistance using QLoRA (Quantized Low-Rank Adaptation).

Model Description

• Base Model: codellama/CodeLlama-7b-hf
• Fine-tuning Method: QLoRA (4-bit quantization with LoRA adapters)
• Framework: Unsloth + Transformers
• Training Data: Custom Python programming examples

Usage

This repository contains LoRA adapters only. To use, merge the adapters with the base model.

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained( model_name="pranav-pvnn/codellama-7b-python-ai-assistant", max_seq_length=2048, load_in_4bit=True, )

prompt = "### Instruction:\nWrite a Python function to calculate factorial.\n### Response:\n" inputs = tokenizer(prompt, return_tensors="pt").to("cuda") outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs, skip_special_tokens=True))

Training Details

• Quantization: 4-bit
• LoRA Rank: 64
• Learning Rate: 2e-4
• Epochs: 4
• Max Seq Length: 2048
• GPU: NVIDIA Tesla T4

Limitations

• Requires base model for inference.
• Optimized for Python code generation.
• Trained on custom dataset (~2,000 examples).

Citation

@misc{codellama-7b-python-assistant, author = {pranav-pvnn}, title = {CodeLlama 7B Python AI Assistant}, year = {2025}, publisher = {HuggingFace}, howpublished = {\url{https://huggingface.co/pranav-pvnn/codellama-7b-python-ai-assistant}} }

Acknowledgements

• Base Model: Meta's CodeLlama
• Training Framework: Unsloth
• Quantization: bitsandbytes
• LoRA: PEFT