IS GPU support enabled for embeddings? Or to put in another way how do I enable GPU for embedding work?

[micos7]

I use the bert model for embeddings. I build it withmake BUILD_TYPE=cublas build. When I test an embedding with a fairly large text it takes the same time as the CPU only build (default), Thanks.

[fakezeta]

Hi @micos7,

transformer backend supports CUDA acceleration for embeddings.