OsaurusAI/gemma-4-E4B-it-qat-JANG_4M

Realign safetensors containers for zero-copy mmap loading (tensor bytes unchanged)

Add files using upload-large-folder tool

initial commit