ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

cuda : add f32 to bf16 copy op (#12806)

Browse source 
This allows BF16 KV-cache on CUDA.
This commit is contained in:
Sigbjørn Skjæret 2025-04-08 23:21:31 +02:00 • committed by GitHub
parent b32efad2bc
commit 7538246e7c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 24 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml/src/ggml-cuda/ggml-cuda.cu
View file

@ -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;
}