Models can be quantized below 100MB with minimal accuracy loss. A full featured Docker image supporting GPU and CPU requires less than 400MB.
確定! 回上一頁