Mosquito

🦟 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

✨ Highlights

🚀 Quick Start

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

🏗️ Architecture