GitHub / intel/torch-xpu-ops issues and pull requests

#1861 - Maxpooling takes too long on BMG

Issue - State: open - Opened by jianyizh 16 days ago

#1860 - Consider removing FP64 patch that forces reference calculations to run on the CPU, since FP64 GEMM is already enabled on the GPU

Issue - State: open - Opened by anmyachev 16 days ago

#1859 - Wrong accuracy of std, var, std_mean, var_mean after update driver from LTS to LTS2.

Issue - State: open - Opened by etaf 16 days ago

#1856 - channel last hardswish will call extra copy

Issue - State: open - Opened by jianyizh 17 days ago

#1854 - Wrong accuracy of aten::prod.dim_int after update driver from LTS to LTS2.

Issue - State: open - Opened by etaf 17 days ago

#1839 - Reduce build warnings in kernel templates

Pull Request - State: closed - Opened by xytintel 22 days ago

#1834 - [release/2.8] Huggingface models inference performance has a little dropped

Issue - State: closed - Opened by mengfei25 22 days ago - 1 comment
Labels: performance, hw: PVC, regression