---
base_model: unsloth/gemma-3-4b-it-unsloth-bnb-4bit
library_name: peft
pipeline_tag: text-generation
tags:
- base_model:adapter:unsloth/gemma-3-4b-it-unsloth-bnb-4bit
- lora
- sft
- transformers
- trl
- unsloth
datasets:
- bitext/Bitext-customer-support-llm-chatbot-training-dataset
---

# Model Card for Model ID

Fine-tuned Gemma 3 4B-IT using QLoRA with 4-bit NF4 quantization and LoRA adapters on the Bitext Customer Support dataset.

## Fine-tuning

- QLoRA
- 4-bit quantization
- Rank 16
- 3 epochs

## Evaluation

| Metric | Base | Fine-tuned |
|--------|------|-----------|
| ROUGE-L | 0.18 | 0.42 |
| BERTScore F1 | 0.843 | 0.913 |

### Qualitative Examples

| User Query | Base Gemma 3 | Fine-tuned Gemma 3 |
|------------|--------------|--------------------|

![image](https://cdn-uploads.huggingface.co/production/uploads/69766193aa07d78946aba56c/ubdCVaq9KKMhpy8d29fdR.png)

## Intended Use

Customer-support assistants for e-commerce and retail domains.