ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-03 13:48:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixup comments

Browse Source
This commit is contained in:
kiefer
2025-09-16 09:07:38 +00:00
parent eea8476738
commit e3fccf0115
62 changed files with 9 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_wmma_cshuffle_v3.hpp
View File

@@ -38,7 +38,7 @@ namespace {
/*
* \brief Wrapper function of GridwiseGemm Wmma Cshuffle V3 to realize grouped forward convolution.
*
* \tparam ComputePtrOffsetOfBatch Class that computes the base pointer offsets of A, B, D and E
* \tparam ComputePtrOffset Class that computes the base pointer offsets of A, B, D and E
* matrices for groups or splitN. Currently it works for identical strides, but this can be extended
* to other layouts. The returned offset can be either \p index_t or \p long_index_t. If it returns
* \p long_index_t, we are not subject to the 2GB limitations.
@@ -46,14 +46,6 @@ namespace {
* \tparam Block2ETileMap Block2ETileMap::CalculateBottomIndex() takes in the id of a workgroup and
* returns the 2D index of the tile that it computes. \see
* GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3::Run().
*
* \note Using \p ComputePtrOffsetOfBatch gives us the flexibility that 2 workgroups can compute 2
* tiles from different matrices. Keep in mind that these 2 matrices can share the same grid
* descriptor (like in BatchedGEMM), or use their own grid descriptors (in GroupedGemm). \link
* impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp kernel_gemm_xdlops_v2r3_for_conv3d \endlink for
* \link DeviceConv3d \endlink uses the same concept, but currently does NOT encapsulate the
* computing of pointer offset into \p ComputePtrOffsetOfStridedBatch.
*
*/
template <typename GridwiseGemm,
typename AGridDesc_AK0_M_AK1,
@@ -124,8 +116,6 @@ __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
#endif // End of if (!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx11__) || defined(__gfx12__))
}
// TODO: Implement 2lds kernel?
} // namespace
template <typename T>
@@ -791,12 +781,6 @@ struct DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3
}
{
// Original effective calculation of MBlock and NBlock
// const auto M = e_grid_desc_m_n.GetLength(I0);
// const auto N = e_grid_desc_m_n.GetLength(I1);
// const auto MBlock = M / MPerBlock;
// const auto NBlock = N / NPerBlock;
const index_t GemmM = a_grid_desc_ak0_m_ak1_.GetLength(I1);
const index_t GemmN = b_grid_desc_bk0_n_bk1_.GetLength(I1);
const auto MBlock = CTranspose ? GridwiseGemmCTranspose::CalculateMBlock(GemmN)
@@ -1149,9 +1133,14 @@ struct DeviceGroupedConvFwdMultipleABD_Wmma_CShuffle_V3
ck::utility::flush_icache();
// rotating mem
rotating_mem.Next();
// clear c mem
// TODO: this E clearing does not look correct. Fix when implementing
// splitK. if(arg_.KBatch > 1)
// clear E mem
// TODO: The calculation of the E buffer size may not be correct in all
// cases, for example if the memory is not contiguous due to padding or
// unusual strides. Investigate when implementing splitK. It may be
// safer to use GetElementSpaceSize().
// if(arg_.KBatch > 1)
// HIP_CHECK_ERROR(hipMemsetAsync(arg_.p_e_grid,
// 0,
// arg_.M * arg_.N *

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/comp/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_f16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_gnhwc_gkyxc_gnhwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instance.in
View File

@@ -21,7 +21,6 @@ using device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instances
PassThrough,
PassThrough>>>;
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
template <int Shards, int ShardIndex>
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instances_shard(
device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_instances& instances)

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_instance.in
View File

@@ -21,7 +21,6 @@ using device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_instances
PassThrough,
PassThrough>>>;
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
template <int Shards, int ShardIndex>
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_instances_shard(
device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_instances& instances)

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_f16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_int8_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkyxc_ngkhw_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_ngchw_gkcyx_ngkhw_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_inter_instance.in
View File

@@ -21,7 +21,6 @@ using device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_inter
PassThrough,
PassThrough>>>;
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
template <int Shards, int ShardIndex>
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_inter_instances_shard(
device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_inter_instances& instances)

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/mem/device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_intra_instance.in
View File

@@ -21,7 +21,6 @@ using device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_intra
PassThrough,
PassThrough>>>;
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
template <int Shards, int ShardIndex>
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_intra_instances_shard(
device_grouped_conv2d_fwd_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_int8_mem_intra_instances& instances)

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/merged_groups/device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_ngchw_gkcyx_ngkhw_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_ngchw_gkcyx_ngkhw_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/merged_groups/device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_ngchw_gkcyx_ngkhw_f16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_ngchw_gkcyx_ngkhw_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NGCHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/merged_groups/device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/merged_groups/device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/merged_groups/device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_int8_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/comp/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/comp/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/mem/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/merged_groups/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_bias_clamp/wmma/merged_groups/device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_bias_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/comp/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/comp/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_comp_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_comp_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_bf16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_mem_inter_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_inter_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/mem/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_fp16_mem_intra_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_nhwgc_gkyxc_nhwgk_f16_mem_intra_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/merged_groups/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd_clamp/wmma/merged_groups/device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_fp16_instance.cpp
View File

@@ -9,7 +9,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv2d_fwd_clamp_wmma_cshufflev3_merged_groups_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC,

1
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instance.in
View File

@@ -7,7 +7,6 @@
namespace ck::tensor_operation::device::instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
using device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_bf16_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_16x16_instance.cpp
View File

@@ -8,7 +8,6 @@ namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
void add_device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_16x16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,

1
library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/wmma/device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_instance.in
View File

@@ -7,7 +7,6 @@
namespace ck::tensor_operation::device::instance {
// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
using device_grouped_conv3d_fwd_wmma_cshufflev3_ngcdhw_gkczyx_ngkdhw_f16_instances =
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
NGCDHW,