Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ggml: add f16 out_prod support for CPU and out_prod op for Vulkan (#23997)

Browse Source
This commit is contained in:
Michael Lamothe
2026-07-15 10:46:56 +02:00
committed by GitHub
parent 33a75f41c3
commit f955e394bf
6 changed files with 167 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/ggml-cpu.cpp
+3 -2
View File
@@ -462,11 +462,12 @@ static bool ggml_backend_cpu_device_supports_op(ggml_backend_dev_t dev, const st
return max_bias == 0.0f;
}
case GGML_OP_IM2COL_BACK:
return src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32;
return src0->type == GGML_TYPE_F32 && (src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16);
case GGML_OP_GET_ROWS_BACK:
return src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16;
case GGML_OP_OUT_PROD:
return (src0->type == GGML_TYPE_F32 || (ggml_is_quantized(src0->type) && src0->ne[2] == src1->ne[2] && src0->ne[3] == src1->ne[3])) &&
return (src0->type == GGML_TYPE_F32 ||
((src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type)) && src0->ne[2] == src1->ne[2] && src0->ne[3] == src1->ne[3])) &&
src1->type == GGML_TYPE_F32 && op->type == GGML_TYPE_F32;
default:
return true;