---
license: mit
tags:
- trl
- sft
datasets:
- yagebin/smart_home_dataset
language:
- en
base_model:
- mistralai/Mistral-7B-Instruct-v0.3
---

# Model Card for Chatbot-Smart-Home

Chatbot-Smart-Home

A fine-tuned **Mistral-7B-Instruct model** to create an intelligent chatbot for **smart home consultations**. This model helps users design a personalized smart home solution by interacting via scenario-based dialogue and recommending devices based on user preferences, property attributes, and budget.

---

## Model Details

### Model Description

This model is tailored to simulate natural conversations for smart home planning. It gathers user preferences, property attributes, and demographic details to recommend smart home devices, calculate costs, and assess ecological ratings.

- **Developed by:** [Ruiqi Tan]
- **Funded by:** Self-funded
- **Shared by:** [Ruiqi Tan]
- **Model type:** Causal Language Model (Fine-tuned Mistral-7B)
- **Language(s) (NLP):** English
- **License:** MIT
- **Finetuned from model:** Mistral-7B-Instruct

### Model Sources

- **Repository:** [yagebin/smart_home_chatbot]


---

## Uses

### Direct Use

This model is designed for:
- Generating scenario-based dialogues to gather user preferences.
- Recommending smart home solutions (e.g., devices, installation metrics, cost).
- Simulating smart home consultations for businesses or personal use.

### Downstream Use

This model can be integrated into:
- Smart home consultation platforms.
- Virtual assistants for smart home system design.
- Research on AI-driven ecological solutions.

### Out-of-Scope Use

This model is not designed for:
- Applications requiring real-time control of smart home devices.
- Use in languages other than English.
- Situations demanding extensive general knowledge outside smart home domains.

---

## Bias, Risks, and Limitations

### Bias
- The model relies on synthesized data, which may introduce biases.
- Limited to scenarios and devices represented in the training data.

### Risks
- The model's responses are deterministic and depend heavily on user inputs, which might lead to inappropriate recommendations if inputs are inconsistent.

### Limitations
- Does not account for regional or cultural differences in smart home preferences.
- Cannot dynamically update to reflect changes in smart device availability or pricing.

### Recommendations
- Validate device recommendations and costs with real-world data.
- Use the model in combination with human oversight for critical use cases.

---

## How to Get Started with the Model

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the model and tokenizer
model_id = "your_username/chatbot-smart-home"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# Example input
input_text = "What smart lighting options can improve energy efficiency?"

# Generate a response
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_length=256)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(response)