OsaurusAI/gemma-4-E4B-it-8bit

Realign model-00002-of-00002.safetensors for zero-copy mmap loading (tensor bytes unchanged)

Realign model-00001-of-00002.safetensors for zero-copy mmap loading (tensor bytes unchanged)

Update usage to Osaurus branding

Gemma 4 E4B-it 8-bit MLX conversion with verified vision+audio weights

initial commit