Xinya Zhang

Results 19 comments of Xinya Zhang

``` bash $ (cd ./branch_build/$(git branch --show-current)/RelWithDebInfo/; ./onnxruntime_test_all --gtest_filter='NGramRepeatBlockTest.*:LongformerAttentionTest.*:DecoderAttentionTest.*:') ... [----------] Global test environment tear-down [==========] 13 tests from 3 test suites ran. (4578 ms total) [ PASSED ] 13...

> pls resolve the conflicts first, thx Updated @ytaous

> The actual impl is not real fused one because MIOpen doesn't have related cudnnConvolutionBiasActivationForward. So I am thinking we just don't enable the ConvActivationFusion on ROCm so we don't...

I'm looking into the Fusion API and trying to implement a real fused version of FusedConv

I just pushed a revised version of FusedConv which uses Fusion API if possible. Due to limitations of MIOpen I kept the old code path as the fall back. For...

I noticed a flaw in previous Fusion API usage, and sent a new commit to address this problem. (See the commit message for the detail) This change is left as...

@ytaous @iK1D @mindest Updated a new version with full features. The only remaining is due to the interface limitation of MIOpen the spatial dimension of convolution operator can only be...

A temporary fix: Replace all `std::string` inside `ompl_util_generator_t.filter_declarations` with `std::basic_string` Update: it doesn't fix all the problems. I'm guessing it's due to some recent updates to castxml. Here is my...

As a reference, this is a known bug in gnome-shell https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/742