nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Disable add + fused_rms_norm fusion (#916) 

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2025-11-07 19:38:18 +02:00 committed by GitHub
parent 1c31b25380
commit d0850dccc8
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda.cu
View File

@ -3102,7 +3102,7 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
ggml_cuda_op_fused_add_add_rms_norm(ctx, dst, cgraph->nodes[i+1], cgraph->nodes[i+2]); ggml_cuda_op_fused_add_add_rms_norm(ctx, dst, cgraph->nodes[i+1], cgraph->nodes[i+2]);
i += 2; i += 2;
} }
else if (fusion && i + 1 < cgraph->n_nodes && else if (false && fusion && i + 1 < cgraph->n_nodes &&
cgraph->nodes[i+1]->op == GGML_OP_FUSED_RMS_NORM && cgraph->nodes[i+1]->op == GGML_OP_FUSED_RMS_NORM &&
ggml_is_contiguous(dst->src[0]) && ggml_is_contiguous(dst->src[0]) &&
ggml_is_contiguous(dst->src[1]) && ggml_is_contiguous(dst->src[1]) &&