Instructions to use Ilides/cortex-0.3-0.02b-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use Ilides/cortex-0.3-0.02b-mlx with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("Ilides/cortex-0.3-0.02b-mlx") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use Ilides/cortex-0.3-0.02b-mlx with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "Ilides/cortex-0.3-0.02b-mlx"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "Ilides/cortex-0.3-0.02b-mlx" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8080/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Ilides/cortex-0.3-0.02b-mlx", "messages": [ {"role": "user", "content": "Hello"} ] }' - Atomic Chat
|
Download README.md from Ilides/cortex-0.3-0.02b-mlx: direct link, hf CLI and curl.
- Browser
- Download file 1.33 kB
-
https://huggingface.co/Ilides/cortex-0.3-0.02b-mlx/resolve/main/README.md
- Command line
-
hf download hf://Ilides/cortex-0.3-0.02b-mlx/README.md
-
curl -L -o README.md https://huggingface.co/Ilides/cortex-0.3-0.02b-mlx/resolve/main/README.md
1.33 kB
| library_name: mlx | |
| pipeline_tag: text-generation | |
| license: apache-2.0 | |
| base_model: Ilides/cortex-0.3-0.02b | |
| tags: | |
| - cortex | |
| - mlx | |
| - apple-silicon | |
| # cortex-0.3-0.02b 路 MLX (Apple Silicon) | |
| Versi贸n **MLX** (Metal) de [Ilides/cortex-0.3-0.02b](https://huggingface.co/Ilides/cortex-0.3-0.02b), el GPT conversacional de 20.6 M par谩metros entrenado desde cero en NumPy puro. Ejecuta nativamente en Apple Silicon sin PyTorch. | |
| ## Requisitos | |
| ```bash | |
| pip install mlx safetensors transformers regex | |
| ``` | |
| ## Uso | |
| ```bash | |
| # bater铆a de prompts de prueba (por defecto) | |
| python generate_mlx.py | |
| # una pregunta | |
| python generate_mlx.py "驴Qu茅 eres?" --max-new 128 | |
| # chat interactivo (escribe "salir" para terminar) | |
| python generate_mlx.py --interactive --temperature 0.8 | |
| ``` | |
| Bundle autocontenido: `weights.safetensors` (45 tensores) + `model.py` (arquitectura) + `generate_mlx.py` (CLI) + tokenizer. No depende de `mlx-lm` ni de c贸digo externo del Hub. | |
| ## Verificaci贸n | |
| Paridad de logits MLX vs NumPy (implementaci贸n de referencia): **max|螖| = 3.65e-05** (< 1e-4). Tokens EOS efectivos: `{0, 6}`. | |
| ## Arquitectura | |
| `cortex`: embeddings de posici贸n aprendidos (sin RoPE), RMSNorm + atenci贸n multi-cabeza + SwiGLU, embeddings atados (`tie_word_embeddings`). Contexto 512, 6 capas, d_model 192, vocab 16384 (BPE byte-level). | |