CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

CANN: weight format to NZ for Ascend310P3 (#14407)

Browse Source 
* weight format to nz for 310p

* remove quant weight format to nz

* clean code

* fix

* make the conditions for converting weights to NZ format consistent

* clean code
This commit is contained in:
chen fan
2025-07-23 11:58:00 +08:00
committed by GitHub
parent 8c988fa41d
commit 14c28dfc50
3 changed files with 118 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
ggml/src/ggml-cann/aclnn_ops.cpp
View File

@@ -1785,8 +1785,27 @@ static void ggml_cann_mat_mul_fp(ggml_backend_cann_context& ctx,
size_t transpose_nb[] = {bcast_weight_nb[1], bcast_weight_nb[0],
bcast_weight_nb[2], bcast_weight_nb[3],
bcast_weight_nb[4], bcast_weight_nb[5]};
aclTensor* acl_weight_tensor =
ggml_cann_create_tensor(weight, transpose_ne, transpose_nb, n_dims);
aclTensor* acl_weight_tensor;
bool weightToNZ = false;
#ifdef ASCEND_310P
weightToNZ = (getenv("GGML_CANN_WEIGHT_NZ") != nullptr);
#endif
if (weightToNZ && is_matmul_weight(weight)) {
int64_t acl_stride[2] = {1, transpose_ne[1]};
// Reverse ne.
std::reverse(transpose_ne, transpose_ne + n_dims);
std::vector<int64_t> storageDims = {transpose_ne[0], transpose_ne[1]};
acl_weight_tensor = aclCreateTensor(
transpose_ne, n_dims, ggml_cann_type_mapping(weight->type), acl_stride,
0, ACL_FORMAT_FRACTAL_NZ, storageDims.data(), 2, weight->data);
} else {
acl_weight_tensor =
ggml_cann_create_tensor(weight, transpose_ne, transpose_nb, n_dims, ACL_FORMAT_ND);
}
aclTensor* acl_dst =
ggml_cann_create_tensor(dst, bcast_dst_ne, bcast_dst_nb, n_dims);