This is faster on big text encoder models than running it on the CPU.
Attach a file by drag & drop or click to upload
