Twin 04

Nemotron-Mini-4B

Family: NVIDIA Nemotron · Parameters: 4B · Ollama tag: nemotron · Source map: nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF · Layers: 32 · Heads: 32 · Hidden: 2048 · Attention: GQA

Tokenizer Embedding 32 Layers 32 Heads GQA Hidden 2048 RMSNorm Head Code

Atom

Attention Mechanism

Implement a grouped-query attention shape check for 32 heads and Nemotron's local prompt budget.

Molecule

Tokenizer

Build a tokenizer probe that records edge-case tokens and prepares the result for JS WORM sealing.

Compound

Embedding Layer

Create an embedding-layer inspector that reports hidden dimension 2048 and emits a sealed handoff payload.

WORM Seal

pending-browser-seal 

      

    

    
WORM loading...