ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama: enable K-shift for quantized KV cache for cuda (#760)

Browse Source 
cuda: add q8_0->f32 cpy operation (#9571)
It will fail on unsupported backends or quant types.

Co-authored-by: Ivan <nekotekina@gmail.com>
This commit is contained in:
firecoperana
2025-09-05 04:54:18 -05:00
committed by GitHub
parent 13c3b6412e
commit 49979ba9e9
2 changed files with 31 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -4119,6 +4119,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q8_0) {
return true;
}
if (src0_type == GGML_TYPE_Q8_0 && src1_type == GGML_TYPE_F32) {
return true;
}
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q4_0) {
return true;
}