TheAaravSikriwal/gemma-2-2b-it-mindcontrol

Gemma 2 2B IT · mindcontrol Q4

4-bit weights of google/gemma-2-2b-it for mindcontrol, an LLM engine written from scratch in Rust and WGSL that runs in the browser on the visitor's own GPU, plus the steering vectors its dials use. Try it at wearechintu.com/mindcontrol.

• model.safetensors: every 2-D weight matrix as <name>.qweight (u32, eight 4-bit codes per word) and <name>.scales (u32, f16 scale | f16 offset << 16) per group of 32; norms stay bf16.
• steering.json: contrastive-activation-addition directions, one per dial.
• config.json, tokenizer.json: unchanged from the base model.

These files are a Model Derivative of Gemma. Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms. Use is also subject to the Gemma Prohibited Use Policy.