Skip to content

GitLab

Commit 761d6d82 authored by danyao12's avatar danyao12
Browse files

hardcode bwd bias reading SrcScalarPerVector

parent 774c9209
Show whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_light_v1.hpp
View file @ 761d6d82
...@@ -1285,7 +1285,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V1 ...@@ -1285,7 +1285,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V1
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector 4, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_light_v2.hpp
View file @ 761d6d82
...@@ -1364,7 +1364,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V2 ...@@ -1364,7 +1364,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_Light_V2
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector 4, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_v1.hpp
View file @ 761d6d82
...@@ -1353,7 +1353,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1 ...@@ -1353,7 +1353,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector 4, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_bwd_xdl_cshuffle_qloop_b2t_v2.hpp
View file @ 761d6d82
...@@ -1409,7 +1409,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V2 ...@@ -1409,7 +1409,7 @@ struct GridwiseBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V2
Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder Sequence<0, 1, 2, 4, 3, 5>, // DstDimAccessOrder
4, // SrcVectorDim 4, // SrcVectorDim
5, // DstVectorDim 5, // DstVectorDim
D0BlockTransferSrcScalarPerVector, // SrcScalarPerVector 4, // SrcScalarPerVector
4, // DstScalarPerVector 4, // DstScalarPerVector
1, 1,
1, 1,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment