klosax

Openblas deafults to some maximum available threads, but would probably not be the most optimal. In Openblas there is a function to set the number of threads, why not use...

System freeze when compiled with cublast

5

When running main compiled with cublast in newest releases (305eb5a), everything works fine until right before returning to command prompt. The timing info pops up then my system completely freeze...

Enabling cuda GPU acceleration

14

It is possible to use cuBLAS by enabling it when compiling: `-DGGML_CUBLAS=ON` Maybe add this to the readme?

Would ggllm.cpp adopt the upcoming ggml fileformat (GGUF)?

1

@cmp-nct PR here: https://github.com/ggerganov/ggml/pull/302

Porting state-of-the-art face recognition deep model to ggml

1

I wonder if the face recognition model used in dlib could be ported to ggml? A port of something like this would open up more possible uses of ggml. https://github.com/davisking/dlib/blob/master/examples/dnn_face_recognition_ex.cpp