frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

[BE]: Update cudnn frontend submodule to 1.15.0 (#165776)

Browse Source 
Update cudnn frontend submodule to 1.15.0
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165776
Approved by: https://github.com/eqy
This commit is contained in:
Aaron Gokaslan
2025-10-18 02:23:22 +00:00
committed by PyTorch MergeBot
parent c137e222d4
commit de09bab4b6
2 changed files with 3 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
aten/src/ATen/native/cudnn/MHA.cpp
View File

@ -487,9 +487,7 @@ std::unique_ptr<fe::graph::Graph> build_graph(
auto scaled_dot_product_flash_attention_options =
fe::graph::SDPA_attributes()
.set_name("CUDNN_SDPA")
.set_is_inference(return_softmaxstats == false)
// TODO(eqy): switch to this API once cuDNN FE is upgraded
// .set_generate_stats(return_softmaxstats)
.set_generate_stats(return_softmaxstats)
.set_causal_mask(is_causal)
.set_attn_scale(attn_scale);
if (use_ragged_in_dense(q, k, v, o, attn_bias.has_value())) {
@ -707,9 +705,7 @@ std::unique_ptr<fe::graph::Graph> build_graph_nestedtensor(
auto scaled_dot_product_flash_attention_options =
fe::graph::SDPA_attributes()
.set_name("CUDNN_SDPA_NESTEDTENSOR")
.set_is_inference(return_softmaxstats == false)
// TODO(eqy): switch to this API once cuDNN FE is upgraded
// .set_generate_stats(return_softmaxstats)
.set_generate_stats(return_softmaxstats)
.set_causal_mask(is_causal)
.set_attn_scale(attn_scale)
.set_seq_len_q(SEQ_LEN_Q_)

2
third_party/cudnn_frontend vendored

Submodule third_party/cudnn_frontend updated: f937055efc...0b1577c8c8