vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cann: fix buffer_num and runtime speed slowly error (#8865)

Browse source
This commit is contained in:
wangshuai09 2024-08-05 21:10:37 +08:00 • committed by GitHub
parent b42978e7e4
commit bc0f887e15
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 15 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml/src/ggml-cann.cpp
View file

@ -1670,10 +1670,6 @@ GGML_CALL static bool ggml_backend_cann_supports_op(ggml_backend_t backend,
// TODO: fix me
// Current groupsize should not be greater than k-1 in
// aclnnWeightQuantBatchMatmulV2GetWorkspaceSize().
if (op->src[0]->ne[0]-1 > QK8_0) {
return true;
}
return false;
case GGML_TYPE_Q4_0:
return true;
default: