mozempk/ggml-org-llama-cpp-mirror
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-11-11 14:22:09 +01:00
Code Issues Packages Projects Releases Wiki Activity

CANN: supports out_prod operator for F32 and F16 (#17406)

Browse Source 
Co-authored-by: tianhao <tianhao42@huawei.com>
This commit is contained in:
TianHao324
2025-11-25 17:39:06 +08:00
committed by GitHub
co-authored by tianhao
parent b1846f1c8e
commit 064c90d843
3 changed files with 95 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cann/ggml-cann.cpp
+13
View File
@@ -1886,6 +1886,9 @@ static bool ggml_cann_compute_forward(ggml_backend_cann_context & ctx, struct gg
case GGML_OP_FLASH_ATTN_EXT:
ggml_cann_flash_attn_ext(ctx, dst);
break;
case GGML_OP_OUT_PROD:
ggml_cann_out_prod(ctx, dst);
break;
default:
return false;
}
@@ -2563,6 +2566,16 @@ static bool ggml_backend_cann_supports_op(ggml_backend_dev_t dev, const ggml_ten
case GGML_OP_PAD_REFLECT_1D:
case GGML_OP_COUNT_EQUAL:
return true;
case GGML_OP_OUT_PROD:
{
switch (op->src[0]->type) {
case GGML_TYPE_F16:
case GGML_TYPE_F32:
return true;
default:
return false;
}
}
case GGML_OP_CONV_TRANSPOSE_1D:
// TODO: ((weightL - 1) * dilationW - padLeft)=1336 should not be larger than 255.
return (op->src[0]->ne[0] - 1) <= 255;