Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[fbgemm_gpu] Build GenAI for ROCm in OSS [WIP] #3415

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[fbgemm_gpu] Build GenAI for ROCm in OSS [WIP] #3415

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion fbgemm_gpu/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -195,7 +195,7 @@ if(NOT FBGEMM_CPU_ONLY)
add_subdirectory(experimental/gemm)
endif()

if(NOT FBGEMM_CPU_ONLY AND NOT USE_ROCM)
if(NOT FBGEMM_CPU_ONLY)
# TODO: Re-enable gen_ai for ROCm once ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_ab_scale.hpp
# lands into latest ROCm
add_subdirectory(experimental/gen_ai)
Expand Down
32 changes: 32 additions & 0 deletions fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/ck_utility.hip
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,8 +23,40 @@
#include "ck/ck.hpp"
#include "ck/stream_config.hpp"
#include "ck/host_utility/hip_check_error.hpp"

#if __has_include("ck/utility/flush_icache.hpp")

#include "ck/utility/flush_icache.hpp"

#else

#include <hip/hip_runtime.h>

namespace ck {
// Copied from: https://github.com/ROCm/composable_kernel/blob/rocm-6.2.0/include/ck/utility/flush_icache.hpp
static __global__ void flush_icache() {
asm __volatile__("s_icache_inv \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t"
"s_nop 0 \n\t" ::
:);
}
} // namespace ck

#endif
namespace fbgemm_gpu {

void flush_icache_ck()
Expand Down
23 changes: 23 additions & 0 deletions fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_blockwise_gemm.hip
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,8 @@
#include <c10/hip/HIPStream.h>
#include <torch/torch.h>

#if (defined(USE_ROCM) && ROCM_VERSION >= 60300)

#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
Expand All @@ -30,6 +32,8 @@
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp"

// NOTE: This source is currently only available in the `develop` branch of CK
// https://github.com/ROCm/composable_kernel
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle_v3_ab_scale.hpp"

// Define commonly used types.
Expand All @@ -40,8 +44,11 @@ using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;

#endif

namespace fbgemm_gpu {

#if (defined(USE_ROCM) && ROCM_VERSION >= 60300)
template <
int BLOCK_SIZE,
int MBLOCK,
Expand Down Expand Up @@ -269,4 +276,20 @@ at::Tensor f8f8bf16_blockwise(
}
}

#else

at::Tensor f8f8bf16_blockwise(
at::Tensor XQ,
at::Tensor WQ,
at::Tensor x_scale,
at::Tensor w_scale,
int64_t block_m = 128,
int64_t block_n = 128,
int64_t block_k = 128) {
throw std::runtime_error(
"ROCm version is older than 6.3"); // requires ROCm>=6.3
}

#endif

} // namespace fbgemm_gpu
Loading