README.md

metadata

title: JuliaGPT-v2
emoji: 🧠
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
license: mit
tags:
  - julia
  - flux-jl
  - gpt2-style
  - philosophy
  - openai-compatible
  - char-level

JuliaGPT-v2 Space

GPT-2 style decoder model (384d, 6L, 6H) trained on classical philosophy. Character-level tokenizer (38 chars). Trained in Julia/Flux.jl, served via PyTorch.

Endpoints

• GET / — Health check and model info
• GET /v1/models — List available models
• POST /v1/chat/completions — Generate text (supports streaming)

Usage

curl -X POST https://LisaMegaWatts-JuliaGPT-v2-space.hf.space/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"messages": [{"role": "user", "content": "the nature of"}], "max_tokens": 200}'

Architecture

• Model: 384d embed, 6 layers, 6 heads, ~4.7M params
• Tokenizer: Character-level (38 chars)
• Normalization: LayerNorm (pre-norm)
• Feed-forward: GELU activation
• Framework: Flux.jl (training) / PyTorch (serving)