|
|
--- |
|
|
license: other |
|
|
license_name: innovatronix-restricted-license |
|
|
license_link: LICENSE |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
--- |
|
|
 |
|
|
# 🏠 Innovatronix Home Automation Language Model (Beta) |
|
|
|
|
|
This model, created by the Innovatronix team, serves as a lightweight LLM model tailored for home automation applications. It is a fine-tuned version of Flan T5, trained on a custom dataset containing data related to home automation. The model is designed for basic conversational interactions and is currently in beta development. |
|
|
|
|
|
## ✨ Features |
|
|
|
|
|
- **Conversational Control:** Engage in dialogue with the model to automate smart home functions, commanding and retrieving data from connected smart devices through natural language interactions. |
|
|
|
|
|
- **Lightweight and Efficient:** Optimized for reduced storage and computational demands, allowing seamless deployment in local environments without excessive resource consumption. |
|
|
|
|
|
- **Versatile Deployment:** Flexibly deployable across various platforms, including mobile applications and web interfaces, providing users with accessibility to control their smart homes from preferred devices. |
|
|
|
|
|
## 📊 Training Data |
|
|
|
|
|
The model was fine-tuned on a meticulously handcrafted dataset encompassing a wide range of commands, queries, and contextual information pertaining to controlling and managing smart devices within a home setting. |
|
|
You can view and download the dataset here: [Dataset](https://drive.google.com/file/d/1O8UvuiTia1SZGHLHdFZ8lHeaeMdeoobw/view?usp=sharing) |
|
|
## ⭐ Example usage |
|
|
|
|
|
```python |
|
|
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("Robin246/inxai_v1.1") |
|
|
model = AutoModelForSeq2SeqLM.from_pretrained("Robin246/inxai_v1.1") |
|
|
|
|
|
# Adjust the parameters if needed |
|
|
def generate_response(input_prompt, model, tokenizer): |
|
|
input_text = f"Input prompt: {input_prompt}" |
|
|
|
|
|
input_ids = tokenizer.encode(input_text, return_tensors="pt", max_length=64, padding="max_length", truncation=True) |
|
|
|
|
|
output_ids = model.generate(input_ids, |
|
|
max_length=256, |
|
|
num_return_sequences=1, |
|
|
num_beams=2, |
|
|
early_stopping=True, |
|
|
#do_sample=True, |
|
|
#temperature=0.8, |
|
|
#top_k=50 |
|
|
) #You can vary the top_k or add any other parameters |
|
|
|
|
|
generated_output = tokenizer.decode(output_ids[0], skip_special_tokens=True) |
|
|
return generated_output |
|
|
|
|
|
|
|
|
while True: |
|
|
user_input = input("Enter prompt: ") |
|
|
user_input = ["{}".format(user_input)] |
|
|
if user_input=='Quit': |
|
|
break |
|
|
else: |
|
|
reply = generate_response(user_input, model, tokenizer) |
|
|
print("Generated Reply({}):".format(model), reply) |
|
|
|
|
|
#INXAI from huggingface 'Robin246/inxai_v1.1' |
|
|
``` |
|
|
### ✎ Citation |
|
|
|
|
|
**Developers:** |
|
|
- Robinkumar |
|
|
- Kiransekar |
|
|
- Magesh |
|
|
- Lathikaa Shri |
|
|
<!-- - Nithishkumar --> |
|
|
|
|
|
**Base model credits** |
|
|
This model was fine-tuned from Google's Flan-T5 Model using a custom dataset. |
