Skip to content

GitLab

Commit fe728dc5 authored by illsilin's avatar illsilin
Browse files

disable instances with blocksize=256 in attention examples

parent 9b69d57f
Hide whitespace changes
Inline Side-by-side
Showing with 6 additions and 6 deletions
example/32_batched_gemm_scale_softmax_gemm/cross_attention_forward_wmma_fp16.cpp
View file @ fe728dc5
...@@ -71,7 +71,7 @@ static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecial ...@@ -71,7 +71,7 @@ static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecial
#define CK_MHA_USE_WAVE_1 #define CK_MHA_USE_WAVE_1
#define CK_MHA_USE_WAVE_2 #define CK_MHA_USE_WAVE_2
#define CK_MHA_USE_WAVE_4 #define CK_MHA_USE_WAVE_4
#define CK_MHA_USE_WAVE_8 //#define CK_MHA_USE_WAVE_8
using DeviceMHAFactory = using DeviceMHAFactory =
std::tuple< std::tuple<
#ifdef CK_MHA_USE_WAVE_1 #ifdef CK_MHA_USE_WAVE_1
...@@ -277,10 +277,10 @@ using DeviceMHAFactory = ...@@ -277,10 +277,10 @@ using DeviceMHAFactory =
S<2, 8, 8>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, false, S<2, 8, 8>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, false,
// CShuffleBlockTransfer MN // CShuffleBlockTransfer MN
1, 1, S<1, 64, 1, 2>, 8, 1, 1, S<1, 64, 1, 2>, 8,
MaskingSpec>, MaskingSpec>
#endif #endif
#ifdef CK_MHA_USE_WAVE_8 #ifdef CK_MHA_USE_WAVE_8
ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle< ,ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle<
NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,
ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType,
AElementOp, B0ElementOp, Acc0ElementOp, B1ElementOp, CElementOp, AElementOp, B0ElementOp, Acc0ElementOp, B1ElementOp, CElementOp,
......
example/32_batched_gemm_scale_softmax_gemm/self_attention_forward_wmma_fp16.cpp
View file @ fe728dc5
...@@ -71,7 +71,7 @@ static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecial ...@@ -71,7 +71,7 @@ static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecial
#define CK_MHA_USE_WAVE_1 #define CK_MHA_USE_WAVE_1
#define CK_MHA_USE_WAVE_2 #define CK_MHA_USE_WAVE_2
#define CK_MHA_USE_WAVE_4 #define CK_MHA_USE_WAVE_4
#define CK_MHA_USE_WAVE_8 //#define CK_MHA_USE_WAVE_8
using DeviceMHAFactory = using DeviceMHAFactory =
std::tuple< std::tuple<
#ifdef CK_MHA_USE_WAVE_1 #ifdef CK_MHA_USE_WAVE_1
...@@ -277,10 +277,10 @@ using DeviceMHAFactory = ...@@ -277,10 +277,10 @@ using DeviceMHAFactory =
S<2, 8, 8>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, false, S<2, 8, 8>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 1, false,
// CShuffleBlockTransfer MN // CShuffleBlockTransfer MN
1, 1, S<1, 64, 1, 2>, 8, 1, 1, S<1, 64, 1, 2>, 8,
MaskingSpec>, MaskingSpec>
#endif #endif
#ifdef CK_MHA_USE_WAVE_8 #ifdef CK_MHA_USE_WAVE_8
ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle< ,ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle<
NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,
ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType,
AElementOp, B0ElementOp, Acc0ElementOp, B1ElementOp, CElementOp, AElementOp, B0ElementOp, Acc0ElementOp, B1ElementOp, CElementOp,
......
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