cuda : add f32 to bf16 copy op (#12806)
This allows BF16 KV-cache on CUDA.
This commit is contained in:
parent
b32efad2bc
commit
7538246e7c
2 changed files with 24 additions and 0 deletions
|
@ -3079,6 +3079,9 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
|
|||
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F32) {
|
||||
return true;
|
||||
}
|
||||
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_BF16) {
|
||||
return true;
|
||||
}
|
||||
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F16) {
|
||||
return true;
|
||||
}
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue