From 4c09ae57bc60c95ec3d0c68698e28b0e16d01bdc Mon Sep 17 00:00:00 2001 From: Enrico Degregori Date: Mon, 6 Oct 2025 12:32:01 +0000 Subject: [PATCH] Disable splitk for 2stage xdl on rdna (bug to be fixed) --- ...device_grouped_conv_bwd_weight_two_stage_xdl_cshuffle.hpp | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_two_stage_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_two_stage_xdl_cshuffle.hpp index 50796f78b4..ed5313af56 100644 --- a/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_two_stage_xdl_cshuffle.hpp +++ b/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_two_stage_xdl_cshuffle.hpp @@ -1745,6 +1745,11 @@ struct DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle { return false; } + // TODO: this is needed because there is a bug + if(arg.k_batch_ > 1) + { + return false; + } } // Check this here, it allows to use other instances from factory even