qwen-lora-chatbot

Qwen LoRA Chatbot

A LoRA fine-tuned version of Qwen/Qwen2.5-0.5B trained on instruction-following data using PEFT and TRL.

Model Details

Developed By

Swayam Saini

Model Type

PEFT / LoRA Adapter for Causal Language Modeling

Base Model

Qwen/Qwen2.5-0.5B

Language

English

Frameworks

• Transformers
• PEFT
• TRL

PEFT Version

0.19.1

Training Dataset

Dataset used:

yahma/alpaca-cleaned

The dataset contains instruction-response pairs designed to improve instruction-following behavior.

Fine-Tuning Configuration

• LoRA Rank (r): 16
• LoRA Alpha: 32
• LoRA Dropout: 0.05
• Precision: FP16
• Trainer: TRL SFTTrainer

Model Architecture

This repository contains a LoRA adapter trained on top of:

Qwen/Qwen2.5-0.5B

The final chatbot is created by combining:

Base Model + LoRA Adapter

Files

• adapter_model.safetensors
• adapter_config.json
• tokenizer.json
• tokenizer_config.json
• chat_template.jinja

Intended Use

This model can be used for:

• Conversational AI
• Chatbots
• Question Answering
• Educational demonstrations
• LLM fine-tuning experiments

Limitations

This model may:

• Hallucinate facts
• Produce inaccurate information
• Reflect biases present in training data

Outputs should be reviewed before use in production systems.

Deployment

The adapter can be loaded using:

from transformers import AutoModelForCausalLM
from peft import PeftModel

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-0.5B"
)

model = PeftModel.from_pretrained(
    base_model,
    "sainiswayam9/qwen-lora-chatbot"
)

Project Summary

This project demonstrates:

• LoRA fine-tuning
• Parameter-efficient training
• Hugging Face Model Hub integration
• Hugging Face Space deployment
• End-to-end LLM workflow

Author

Swayam Saini