CoolFace
Apppublic

fourlhs/nano-gpt-z

sourceHugging Faceupdated 6mo agoView on Hugging Face
0likes
App README

NanoGPT

A lightweight Gen Z slang language model powered by a finetuned GPT-2 architecture, trained on 1B tokens.

Features

  • —Fast inference: ~3-5 seconds per generation on free GPU
  • —Gen Z slang: Trained specifically on internet slang and colloquial language
  • —Simple interface: Clean Gradio chat UI
  • —Adjustable parameters: Control temperature and top-p sampling

Usage

Type any prompt and the model will generate a response in Gen Z slang style.

Parameters

  • —Temperature: Controls randomness (0.1-1.5). Higher = more creative, lower = more deterministic
  • —Top P: Nucleus sampling threshold (0.1-1.0). Controls diversity of token selection

Model Details

  • —Architecture: GPT-2 style transformer
  • —Vocabulary: 50,257 tokens (GPT-2 BPE encoding)
  • —Training: 1B tokens of Gen Z slang data
  • —Checkpoint: finetune_genz_1000k_best.pt

How to Run Locally

bash
# Install dependencies
pip install -r hf_spaces/requirements.txt

# Run the app
python hf_spaces/app.py

The app will be available at http://localhost:7860

Deployment

This Space uses Docker for deployment on HuggingFace Spaces.