Remove all imports of <torch/torch.h> from CK extensions (pytorch#4407)

Summary:
Pull Request resolved: pytorch#4407

X-link: facebookresearch/FBGEMM#1477

As we are starting to integrate FBGEMM GenAI into torch, we ran into this issue.

The problem is that we cannot depend on the libtorch include during the build. It's actually fine to use other things from their own import. In this case, I think we don't actually need it at all?

Reviewed By: samanamp

Differential Revision: D77393016

fbshipit-source-id: 25e14e7fe4b6c6b9fc1b5c67bad593effbc5ead8

Author: cthi

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/bf16_grouped/bf16_grouped_gemm.hip

@@ -18,7 +18,6 @@
 
 #include <ATen/ATen.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_xdl_cshuffle_tile_loop.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/bf16_grouped/kernels/bf16_grouped_common.h

@@ -12,7 +12,6 @@
 #else
 #include <c10/cuda/CUDAStream.h>
 #endif
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/ck_utility.hip

@@ -16,7 +16,6 @@
 
 #include <ATen/ATen.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #if defined(USE_ROCM)
 

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_blockwise_gemm.hip

@@ -13,7 +13,6 @@
 
 #include <ATen/ATen.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise/fp8_rowwise_gemm.hip

@@ -16,7 +16,6 @@
 
 #include <ATen/ATen.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #include "kernels/fp8_rowwise_kernel_manifest.h"
 

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise/kernels/fp8_rowwise_common.h

@@ -14,7 +14,6 @@
 #else
 #include <c10/cuda/CUDAStream.h>
 #endif
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_batched/fp8_rowwise_batched_gemm.hip

@@ -16,7 +16,6 @@
 
 #include <ATen/ATen.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #include "kernels/fp8_rowwise_batched_kernel_manifest.h"
 

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_batched/kernels/fp8_rowwise_batched_common.h

@@ -12,7 +12,6 @@
 #else
 #include <c10/cuda/CUDAStream.h>
 #endif
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_grouped/fp8_rowwise_grouped_gemm.hip

@@ -19,7 +19,6 @@
 #include <ATen/ATen.h>
 #include <ATen/hip/HIPContext.h>
 #include <c10/hip/HIPStream.h>
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_xdl_cshuffle_tile_loop.hpp"

fbgemm_gpu/experimental/gen_ai/src/quantize/ck_extensions/fp8_rowwise_grouped/kernels/fp8_rowwise_grouped_common.h

@@ -12,7 +12,6 @@
 #else
 #include <c10/cuda/CUDAStream.h>
 #endif
-#include <torch/torch.h>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"