|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
pipeline_tag: text2text-generation |
|
|
widget: |
|
|
- text: "question: Why is the sky blue?" |
|
|
- text: "question: How do vaccines work?" |
|
|
- text: "question: What is gravity?" |
|
|
- text: "question: How do bees make honey?" |
|
|
- text: "question: What causes earthquakes?" |
|
|
spaces: |
|
|
- ag14850/Mosquito-Demo |
|
|
tags: |
|
|
- t5 |
|
|
- question-answering |
|
|
- knowledge |
|
|
- tiny |
|
|
- efficient |
|
|
- edge |
|
|
- mobile |
|
|
- distillation |
|
|
- iot |
|
|
base_model: |
|
|
- google/t5-v1_1-base |
|
|
--- |
|
|
|
|
|
# 🦟 Mosquito - Tiny Knowledge Model |
|
|
|
|
|
A **7.3M parameter** T5-based model that answers general knowledge questions. Smaller than a mosquito's brain, but surprisingly capable! |
|
|
|
|
|
**🎮 [Try the Live Demo](https://huggingface.co/spaces/ag14850/Mosquito-Demo)** |
|
|
|
|
|
## ✨ Highlights |
|
|
|
|
|
| Feature | Value | |
|
|
|---------|-------| |
|
|
| Parameters | 7,263,744 | |
|
|
| Architecture | T5 v1.1 (Gated FFN) | |
|
|
| Size (FP32) | 29 MB | |
|
|
| Size (quantized + compressed) | ~6 MB | |
|
|
| Training | Knowledge distillation | |
|
|
| License | Apache 2.0 | |
|
|
|
|
|
## 🚀 Quick Start |
|
|
```python |
|
|
from transformers import T5ForConditionalGeneration, AutoTokenizer |
|
|
|
|
|
model = T5ForConditionalGeneration.from_pretrained("ag14850/Mosquito") |
|
|
tokenizer = AutoTokenizer.from_pretrained("google/t5-v1_1-base", legacy=False) |
|
|
|
|
|
def ask(question): |
|
|
inputs = tokenizer(f"question: {question}", return_tensors="pt", max_length=128, truncation=True) |
|
|
outputs = model.generate( |
|
|
**inputs, |
|
|
max_new_tokens=24, |
|
|
num_beams=6, |
|
|
no_repeat_ngram_size=2, |
|
|
repetition_penalty=20.0, |
|
|
early_stopping=True |
|
|
) |
|
|
return tokenizer.decode(outputs[0], skip_special_tokens=True) |
|
|
|
|
|
print(ask("Why is the sky blue?")) |
|
|
# Rayleigh scattering scatters blue wavelengths, causing blue wavelengths. |
|
|
``` |
|
|
|
|
|
## 📊 Example Outputs |
|
|
|
|
|
| Question | Answer | |
|
|
|----------|--------| |
|
|
| Why is the sky blue? | Rayleigh scattering scatters blue wavelengths, causing blue wavelengths. | |
|
|
| How do vaccines work? | Vaccines stimulate the immune system to recognize and fight specific pathogens. | |
|
|
| What is gravity? | Gravity is the force exerted by a planet's gravitational pull. | |
|
|
| How does photosynthesis work? | Photosynthesis converts light energy into chemical energy through photosynthesis. | |
|
|
| What causes earthquakes? | Earthquakes occur when tectonic plates slide past each other, causing earthquakes. | |
|
|
| How do bees make honey? | Bees collect nectar from flowers, converting it into honey. | |
|
|
| What is democracy? | Democracy is a form of government in which citizens vote for representatives. | |
|
|
| Why do we dream? | Dreams are thought to help process emotions and consolidate memories. | |
|
|
|
|
|
## 🏗️ Architecture |
