142 lines
4.0 KiB
Markdown
142 lines
4.0 KiB
Markdown
---
|
|
license: other
|
|
license_name: qwen
|
|
license_link: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct/blob/main/LICENSE
|
|
base_model: unsloth/Qwen3-4B-Instruct-2507
|
|
tags:
|
|
- customer-support
|
|
- chatbot
|
|
- conversational-ai
|
|
- tool-calling
|
|
- entity-extraction
|
|
- unsloth
|
|
- qwen
|
|
- trl
|
|
- sft
|
|
language:
|
|
- en
|
|
- multilingual
|
|
datasets:
|
|
- bitext/Bitext-customer-support-llm-chatbot-training-dataset
|
|
library_name: transformers
|
|
pipeline_tag: text-generation
|
|
model-index:
|
|
- name: Qwen3-4B-customer-support
|
|
results: []
|
|
---
|
|
# Qwen3-4B Customer Support Fine-Tuned Model
|
|
|
|
This is a fine-tuned version of [Qwen3-4B-customer-support](https://huggingface.co/ragib01/Qwen3-4B-customer-support) specifically optimized for customer support interactions. The model has been trained to handle common customer service scenarios including order tracking, refunds, invoice management, and general inquiries.
|
|
|
|
<a href="https://chat.qwen.ai" target="_blank" style="margin: 2px;">
|
|
<img alt="Chat" src="https://img.shields.io/badge/%F0%9F%92%9C%EF%B8%8F%20Qwen%20Chat%20-536af5" style="display: inline-block; vertical-align: middle;"/>
|
|
</a>
|
|
|
|
## Model Description
|
|
|
|
- **Base Model:** unsloth/Qwen3-4B-Instruct-2507
|
|
|
|
- **Fine-tuning Method:** QLoRA (4-bit quantization with LoRA adapters)
|
|
|
|
- **Training Framework:** Unsloth + TRL
|
|
|
|
- **Parameters:** 4B (4,055,498,240 total parameters)
|
|
|
|
- **Trainable Parameters:** 33,030,144 (0.81% of total)
|
|
|
|
- **Language Support:** English + Multilingual capabilities from base model
|
|
|
|
## Key Features
|
|
|
|
✅ **Tool-Calling Capability** - Trained to use structured tool calls for data retrieval (order tracking, invoice lookup, refund processing)
|
|
|
|
✅ **Entity Extraction** - Accurately extracts and preserves values like order numbers, dates, email addresses, and account information
|
|
|
|
✅ **Multilingual Support** - Inherits multilingual capabilities from Qwen3 base model
|
|
|
|
✅ **Memory Efficient** - Trained with 4-bit quantization and LoRA adapters
|
|
|
|
✅ **MCP Compatible** - Architecture preserved for Model Context Protocol compatibility
|
|
|
|
|
|
## Usage
|
|
|
|
### Basic Inference
|
|
The following contains a code snippet illustrating how to use the model generate content based on given inputs.
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
import torch
|
|
|
|
# Load model
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|
"ragib01/Qwen3-4B-customer-support",
|
|
torch_dtype=torch.float16,
|
|
device_map="auto",
|
|
trust_remote_code=True
|
|
)
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(
|
|
"ragib01/Qwen3-4B-customer-support",
|
|
trust_remote_code=True
|
|
)
|
|
|
|
# Test with a customer support query
|
|
messages = [
|
|
{"role": "system", "content": "You are a helpful customer support assistant."},
|
|
{"role": "user", "content": "How do I track my order #74758657?"}
|
|
]
|
|
|
|
# Format input
|
|
text = tokenizer.apply_chat_template(
|
|
messages,
|
|
tokenize=False,
|
|
add_generation_prompt=True
|
|
)
|
|
|
|
# Generate response
|
|
inputs = tokenizer(text, return_tensors="pt").to(model.device)
|
|
|
|
with torch.no_grad():
|
|
outputs = model.generate(
|
|
**inputs,
|
|
max_new_tokens=256,
|
|
temperature=0.7,
|
|
top_p=0.9,
|
|
do_sample=True
|
|
)
|
|
|
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
print(response)
|
|
```
|
|
|
|
### Tool-Calling Support
|
|
|
|
The model can generate structured tool calls for actions requiring data retrieval:
|
|
|
|
```python
|
|
# Example: The model will generate tool calls for order tracking
|
|
|
|
user_query = "Can you check the status of order #98765432?"
|
|
|
|
# Model output will include:
|
|
|
|
<tool_call>
|
|
{
|
|
"name": "track_order",
|
|
"arguments": {
|
|
"order_number": "#98765432"
|
|
}
|
|
}
|
|
</tool_call>
|
|
|
|
```
|
|
|
|
## Use Cases
|
|
|
|
- **Customer Support Chatbots** - Automated responses for common inquiries
|
|
- **Order Management** - Track orders, cancel orders, modify shipping details
|
|
- **Refund Processing** - Handle refund requests and track refund status
|
|
- **Invoice Management** - Retrieve and explain invoice details
|
|
- **Account Management** - Help with account-related questions
|
|
- **Product Information** - Answer questions about products, shipping, and policies |