Use fp16 text encoder weights for CPU inference to lower memory usage.
Attach a file by drag & drop or click to upload
