Add files using upload-large-folder tool

.gitattributes

@@ -119,3 +119,6 @@ phivenv/Lib/site-packages/torch/lib/libprotobuf-lite.lib filter=lfs diff=lfs mer
 phivenv/Lib/site-packages/torch/lib/kineto.lib filter=lfs diff=lfs merge=lfs -text
 phivenv/Lib/site-packages/torch/lib/libprotobuf.lib filter=lfs diff=lfs merge=lfs -text
 phivenv/Lib/site-packages/torch/lib/libprotoc.lib filter=lfs diff=lfs merge=lfs -text
+phivenv/Lib/site-packages/torch/lib/pthreadpool.lib filter=lfs diff=lfs merge=lfs -text
+phivenv/Lib/site-packages/torch/lib/microkernels-prod.lib filter=lfs diff=lfs merge=lfs -text
+phivenv/Lib/site-packages/torch/lib/sleef.lib filter=lfs diff=lfs merge=lfs -text

phivenv/Lib/site-packages/torch/lib/microkernels-prod.lib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d40e51bd2bb25f60758fce200d2eea78bba5eb96e9771bec4d803ec6786ec11d
+size 20267954

phivenv/Lib/site-packages/torch/lib/pthreadpool.lib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88e2da3e5e79047835a21d009707ea855b72d9dc78e46f179886a41f075f75c8
+size 768704

phivenv/Lib/site-packages/torch/lib/sleef.lib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1164e5cee1a9bd4621704221b15f8a24ea1536bc5cf267579c3aa0397d539bbc
+size 8776772

phivenv/Lib/site-packages/transformers/kernels/deta/cpu/ms_deform_attn_cpu.cpp

@@ -0,0 +1,40 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#include <vector>
+
+#include <ATen/ATen.h>
+#include <ATen/cuda/CUDAContext.h>
+
+
+at::Tensor
+ms_deform_attn_cpu_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step)
+{
+    AT_ERROR("Not implement on cpu");
+}
+
+std::vector<at::Tensor>
+ms_deform_attn_cpu_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step)
+{
+    AT_ERROR("Not implement on cpu");
+}

phivenv/Lib/site-packages/transformers/kernels/deta/cpu/ms_deform_attn_cpu.h

@@ -0,0 +1,32 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#pragma once
+#include <torch/extension.h>
+
+at::Tensor
+ms_deform_attn_cpu_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step);
+
+std::vector<at::Tensor>
+ms_deform_attn_cpu_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step);
+

phivenv/Lib/site-packages/transformers/kernels/deta/cuda/ms_deform_attn_cuda.cu

@@ -0,0 +1,156 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#include <vector>
+#include "cuda/ms_deform_im2col_cuda.cuh"
+
+#include <ATen/ATen.h>
+#include <ATen/cuda/CUDAContext.h>
+#include <cuda.h>
+#include <cuda_runtime.h>
+
+#pragma once
+#include <torch/extension.h>
+
+
+at::Tensor ms_deform_attn_cuda_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step)
+{
+    AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
+    AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
+    AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
+    AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
+    AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
+
+    AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
+    AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
+    AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
+    AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
+    AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
+
+    const int batch = value.size(0);
+    const int spatial_size = value.size(1);
+    const int num_heads = value.size(2);
+    const int channels = value.size(3);
+
+    const int num_levels = spatial_shapes.size(0);
+
+    const int num_query = sampling_loc.size(1);
+    const int num_point = sampling_loc.size(4);
+
+    const int im2col_step_ = std::min(batch, im2col_step);
+
+    AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
+    
+    auto output = at::zeros({batch, num_query, num_heads, channels}, value.options());
+
+    const int batch_n = im2col_step_;
+    auto output_n = output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
+    auto per_value_size = spatial_size * num_heads * channels;
+    auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
+    auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
+    for (int n = 0; n < batch/im2col_step_; ++n)
+    {
+        auto columns = output_n.select(0, n);
+        AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_forward_cuda", ([&] {
+            ms_deformable_im2col_cuda(at::cuda::getCurrentCUDAStream(),
+                value.data<scalar_t>() + n * im2col_step_ * per_value_size,
+                spatial_shapes.data<int64_t>(),
+                level_start_index.data<int64_t>(),
+                sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
+                batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
+                columns.data<scalar_t>());
+
+        }));
+    }
+
+    output = output.view({batch, num_query, num_heads*channels});
+
+    return output;
+}
+
+
+std::vector<at::Tensor> ms_deform_attn_cuda_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step)
+{
+
+    AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
+    AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
+    AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
+    AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
+    AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
+    AT_ASSERTM(grad_output.is_contiguous(), "grad_output tensor has to be contiguous");
+
+    AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
+    AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
+    AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
+    AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
+    AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
+    AT_ASSERTM(grad_output.type().is_cuda(), "grad_output must be a CUDA tensor");
+
+    const int batch = value.size(0);
+    const int spatial_size = value.size(1);
+    const int num_heads = value.size(2);
+    const int channels = value.size(3);
+
+    const int num_levels = spatial_shapes.size(0);
+
+    const int num_query = sampling_loc.size(1);
+    const int num_point = sampling_loc.size(4);
+
+    const int im2col_step_ = std::min(batch, im2col_step);
+
+    AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
+
+    auto grad_value = at::zeros_like(value);
+    auto grad_sampling_loc = at::zeros_like(sampling_loc);
+    auto grad_attn_weight = at::zeros_like(attn_weight);
+
+    const int batch_n = im2col_step_;
+    auto per_value_size = spatial_size * num_heads * channels;
+    auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
+    auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
+    auto grad_output_n = grad_output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
+    
+    for (int n = 0; n < batch/im2col_step_; ++n)
+    {
+        auto grad_output_g = grad_output_n.select(0, n);
+        AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_backward_cuda", ([&] {
+            ms_deformable_col2im_cuda(at::cuda::getCurrentCUDAStream(),
+                                    grad_output_g.data<scalar_t>(),
+                                    value.data<scalar_t>() + n * im2col_step_ * per_value_size,
+                                    spatial_shapes.data<int64_t>(),
+                                    level_start_index.data<int64_t>(),
+                                    sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                                    attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
+                                    batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
+                                    grad_value.data<scalar_t>() +  n * im2col_step_ * per_value_size,
+                                    grad_sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                                    grad_attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size);
+
+        }));
+    }
+
+    return {
+        grad_value, grad_sampling_loc, grad_attn_weight
+    };
+}

phivenv/Lib/site-packages/transformers/kernels/deta/cuda/ms_deform_attn_cuda.cuh

@@ -0,0 +1,1467 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#include <vector>
+
+#include <cuda.h>
+#include <cuda_runtime.h>
+
+#include <cstdio>
+#include <algorithm>
+#include <cstring>
+
+#include <ATen/ATen.h>
+#include <ATen/cuda/CUDAContext.h>
+
+#include <THC/THCAtomics.cuh>
+
+#define CUDA_KERNEL_LOOP(i, n)                          \
+  for (int i = blockIdx.x * blockDim.x + threadIdx.x;   \
+      i < (n);                                          \
+      i += blockDim.x * gridDim.x)
+
+
+at::Tensor ms_deform_attn_cuda_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step)
+{
+    AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
+    AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
+    AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
+    AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
+    AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
+
+    AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
+    AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
+    AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
+    AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
+    AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
+
+    const int batch = value.size(0);
+    const int spatial_size = value.size(1);
+    const int num_heads = value.size(2);
+    const int channels = value.size(3);
+
+    const int num_levels = spatial_shapes.size(0);
+
+    const int num_query = sampling_loc.size(1);
+    const int num_point = sampling_loc.size(4);
+
+    const int im2col_step_ = std::min(batch, im2col_step);
+
+    AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
+    
+    auto output = at::zeros({batch, num_query, num_heads, channels}, value.options());
+
+    const int batch_n = im2col_step_;
+    auto output_n = output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
+    auto per_value_size = spatial_size * num_heads * channels;
+    auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
+    auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
+    for (int n = 0; n < batch/im2col_step_; ++n)
+    {
+        auto columns = output_n.select(0, n);
+        AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_forward_cuda", ([&] {
+            ms_deformable_im2col_cuda(at::cuda::getCurrentCUDAStream(),
+                value.data<scalar_t>() + n * im2col_step_ * per_value_size,
+                spatial_shapes.data<int64_t>(),
+                level_start_index.data<int64_t>(),
+                sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
+                batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
+                columns.data<scalar_t>());
+
+        }));
+    }
+
+    output = output.view({batch, num_query, num_heads*channels});
+
+    return output;
+}
+
+
+std::vector<at::Tensor> ms_deform_attn_cuda_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step)
+{
+
+    AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
+    AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
+    AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
+    AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
+    AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
+    AT_ASSERTM(grad_output.is_contiguous(), "grad_output tensor has to be contiguous");
+
+    AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
+    AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
+    AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
+    AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
+    AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
+    AT_ASSERTM(grad_output.type().is_cuda(), "grad_output must be a CUDA tensor");
+
+    const int batch = value.size(0);
+    const int spatial_size = value.size(1);
+    const int num_heads = value.size(2);
+    const int channels = value.size(3);
+
+    const int num_levels = spatial_shapes.size(0);
+
+    const int num_query = sampling_loc.size(1);
+    const int num_point = sampling_loc.size(4);
+
+    const int im2col_step_ = std::min(batch, im2col_step);
+
+    AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
+
+    auto grad_value = at::zeros_like(value);
+    auto grad_sampling_loc = at::zeros_like(sampling_loc);
+    auto grad_attn_weight = at::zeros_like(attn_weight);
+
+    const int batch_n = im2col_step_;
+    auto per_value_size = spatial_size * num_heads * channels;
+    auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
+    auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
+    auto grad_output_n = grad_output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
+    
+    for (int n = 0; n < batch/im2col_step_; ++n)
+    {
+        auto grad_output_g = grad_output_n.select(0, n);
+        AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_backward_cuda", ([&] {
+            ms_deformable_col2im_cuda(at::cuda::getCurrentCUDAStream(),
+                                    grad_output_g.data<scalar_t>(),
+                                    value.data<scalar_t>() + n * im2col_step_ * per_value_size,
+                                    spatial_shapes.data<int64_t>(),
+                                    level_start_index.data<int64_t>(),
+                                    sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                                    attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
+                                    batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
+                                    grad_value.data<scalar_t>() +  n * im2col_step_ * per_value_size,
+                                    grad_sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
+                                    grad_attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size);
+
+        }));
+    }
+
+    return {
+        grad_value, grad_sampling_loc, grad_attn_weight
+    };
+}
+
+const int CUDA_NUM_THREADS = 1024;
+inline int GET_BLOCKS(const int N, const int num_threads)
+{
+  return (N + num_threads - 1) / num_threads;
+}
+
+
+template <typename scalar_t>
+__device__ scalar_t ms_deform_attn_im2col_bilinear(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+  }
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  return val;
+}
+
+
+template <typename scalar_t>
+__device__ void ms_deform_attn_col2im_bilinear(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c,
+                                                   const scalar_t &top_grad,
+                                                   const scalar_t &attn_weight,
+                                                   scalar_t* &grad_value, 
+                                                   scalar_t* grad_sampling_loc,
+                                                   scalar_t* grad_attn_weight)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+  const scalar_t top_grad_value = top_grad * attn_weight;
+  scalar_t grad_h_weight = 0, grad_w_weight = 0;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+    grad_h_weight -= hw * v1;
+    grad_w_weight -= hh * v1;
+    atomicAdd(grad_value+ptr1, w1*top_grad_value);
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+    grad_h_weight -= lw * v2;
+    grad_w_weight += hh * v2;
+    atomicAdd(grad_value+ptr2, w2*top_grad_value);
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+    grad_h_weight += hw * v3;
+    grad_w_weight -= lh * v3;
+    atomicAdd(grad_value+ptr3, w3*top_grad_value); 
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+    grad_h_weight += lw * v4;
+    grad_w_weight += lh * v4;
+    atomicAdd(grad_value+ptr4, w4*top_grad_value);
+  }
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  *grad_attn_weight = top_grad * val;
+  *grad_sampling_loc = width * grad_w_weight * top_grad_value;
+  *(grad_sampling_loc + 1) = height * grad_h_weight * top_grad_value;
+}
+
+
+template <typename scalar_t>
+__device__ void ms_deform_attn_col2im_bilinear_gm(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c,
+                                                   const scalar_t &top_grad,
+                                                   const scalar_t &attn_weight,
+                                                   scalar_t* &grad_value, 
+                                                   scalar_t* grad_sampling_loc,
+                                                   scalar_t* grad_attn_weight)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+  const scalar_t top_grad_value = top_grad * attn_weight;
+  scalar_t grad_h_weight = 0, grad_w_weight = 0;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+    grad_h_weight -= hw * v1;
+    grad_w_weight -= hh * v1;
+    atomicAdd(grad_value+ptr1, w1*top_grad_value);
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+    grad_h_weight -= lw * v2;
+    grad_w_weight += hh * v2;
+    atomicAdd(grad_value+ptr2, w2*top_grad_value);
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+    grad_h_weight += hw * v3;
+    grad_w_weight -= lh * v3;
+    atomicAdd(grad_value+ptr3, w3*top_grad_value); 
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+    grad_h_weight += lw * v4;
+    grad_w_weight += lh * v4;
+    atomicAdd(grad_value+ptr4, w4*top_grad_value);
+  }
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  atomicAdd(grad_attn_weight, top_grad * val); 
+  atomicAdd(grad_sampling_loc, width * grad_w_weight * top_grad_value);
+  atomicAdd(grad_sampling_loc + 1, height * grad_h_weight * top_grad_value);
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_im2col_gpu_kernel(const int n,
+                                                const scalar_t *data_value, 
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *data_col)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    scalar_t *data_col_ptr = data_col + index;
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+    scalar_t col = 0;
+    
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const scalar_t *data_value_ptr = data_value + (data_value_ptr_init_offset + level_start_id * qid_stride);
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          col += ms_deform_attn_im2col_bilinear(data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col) * weight;
+        }
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+      }
+    }
+    *data_col_ptr = col;
+  }
+}
+
+template <typename scalar_t, unsigned int blockSize>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2];
+    __shared__ scalar_t cache_grad_attn_weight[blockSize];
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+        if (tid == 0)
+        {
+          scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0];
+          int sid=2;
+          for (unsigned int tid = 1; tid < blockSize; ++tid)
+          {
+            _grad_w += cache_grad_sampling_loc[sid];
+            _grad_h += cache_grad_sampling_loc[sid + 1];
+            _grad_a += cache_grad_attn_weight[tid];
+            sid += 2;
+          }
+          
+          
+          *grad_sampling_loc = _grad_w;
+          *(grad_sampling_loc + 1) = _grad_h;
+          *grad_attn_weight = _grad_a;
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t, unsigned int blockSize>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2];
+    __shared__ scalar_t cache_grad_attn_weight[blockSize];
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockSize/2; s>0; s>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        { 
+          *grad_sampling_loc = cache_grad_sampling_loc[0];
+          *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1];
+          *grad_attn_weight = cache_grad_attn_weight[0];
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v1(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+        if (tid == 0)
+        {
+          scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0];
+          int sid=2;
+          for (unsigned int tid = 1; tid < blockDim.x; ++tid)
+          {
+            _grad_w += cache_grad_sampling_loc[sid];
+            _grad_h += cache_grad_sampling_loc[sid + 1];
+            _grad_a += cache_grad_attn_weight[tid];
+            sid += 2;
+          }
+          
+          
+          *grad_sampling_loc = _grad_w;
+          *(grad_sampling_loc + 1) = _grad_h;
+          *grad_attn_weight = _grad_a;
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+            if (tid + (s << 1) < spre)
+            {
+              cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)];
+              cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)];
+              cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)];
+            } 
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        {
+          *grad_sampling_loc = cache_grad_sampling_loc[0];
+          *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1];
+          *grad_attn_weight = cache_grad_attn_weight[0];
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+            if (tid + (s << 1) < spre)
+            {
+              cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)];
+              cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)];
+              cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)];
+            }
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        {
+          atomicAdd(grad_sampling_loc, cache_grad_sampling_loc[0]);
+          atomicAdd(grad_sampling_loc + 1, cache_grad_sampling_loc[1]);
+          atomicAdd(grad_attn_weight, cache_grad_attn_weight[0]);
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_gm(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear_gm(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            grad_sampling_loc, grad_attn_weight);
+        }
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+void ms_deformable_im2col_cuda(cudaStream_t stream,
+                              const scalar_t* data_value,
+                              const int64_t* data_spatial_shapes, 
+                              const int64_t* data_level_start_index, 
+                              const scalar_t* data_sampling_loc,
+                              const scalar_t* data_attn_weight,
+                              const int batch_size,
+                              const int spatial_size, 
+                              const int num_heads, 
+                              const int channels, 
+                              const int num_levels, 
+                              const int num_query,
+                              const int num_point,
+                              scalar_t* data_col)
+{
+  const int num_kernels = batch_size * num_query * num_heads * channels;
+  const int num_actual_kernels = batch_size * num_query * num_heads * channels;
+  const int num_threads = CUDA_NUM_THREADS;
+  ms_deformable_im2col_gpu_kernel<scalar_t>
+      <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+          0, stream>>>(
+      num_kernels, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, 
+      batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, data_col);
+  
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess)
+  {
+    printf("error in ms_deformable_im2col_cuda: %s\n", cudaGetErrorString(err));
+  }
+
+}
+
+template <typename scalar_t>
+void ms_deformable_col2im_cuda(cudaStream_t stream,
+                              const scalar_t* grad_col,
+                              const scalar_t* data_value,
+                              const int64_t * data_spatial_shapes,
+                              const int64_t * data_level_start_index,
+                              const scalar_t * data_sampling_loc,
+                              const scalar_t * data_attn_weight,
+                              const int batch_size, 
+                              const int spatial_size, 
+                              const int num_heads,
+                              const int channels, 
+                              const int num_levels,
+                              const int num_query,
+                              const int num_point, 
+                              scalar_t* grad_value,
+                              scalar_t* grad_sampling_loc,
+                              scalar_t* grad_attn_weight)
+{
+  const int num_threads = (channels > CUDA_NUM_THREADS)?CUDA_NUM_THREADS:channels;
+  const int num_kernels = batch_size * num_query * num_heads * channels;
+  const int num_actual_kernels = batch_size * num_query * num_heads * channels;
+  if (channels > 1024)
+  {
+    if ((channels & 1023) == 0)
+    {
+      ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+    }
+    else
+    {
+      ms_deformable_col2im_gpu_kernel_gm<scalar_t>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+    }
+  }
+  else{
+    switch(channels)
+    {
+      case 1:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 1>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 2:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 2>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 4:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 4>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 8:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 8>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 16:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 16>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 32:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 32>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 64:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 64>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 128:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 128>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 256:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 256>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 512:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 512>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 1024:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 1024>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      default:
+        if (channels < 64)
+        {
+          ms_deformable_col2im_gpu_kernel_shm_reduce_v1<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+        }
+        else
+        {
+          ms_deformable_col2im_gpu_kernel_shm_reduce_v2<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+        }
+    }
+  }
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess)
+  {
+    printf("error in ms_deformable_col2im_cuda: %s\n", cudaGetErrorString(err));
+  }
+
+}

phivenv/Lib/site-packages/transformers/kernels/deta/cuda/ms_deform_attn_cuda.h

@@ -0,0 +1,29 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#pragma once
+#include <torch/extension.h>
+
+at::Tensor ms_deform_attn_cuda_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step);
+
+std::vector<at::Tensor> ms_deform_attn_cuda_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step);

phivenv/Lib/site-packages/transformers/kernels/deta/cuda/ms_deform_im2col_cuda.cuh

@@ -0,0 +1,1327 @@
+/*!
+**************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************
+* Modified from DCN (https://github.com/msracver/Deformable-ConvNets)
+* Copyright (c) 2018 Microsoft
+**************************************************************************
+*/
+
+#include <cstdio>
+#include <algorithm>
+#include <cstring>
+
+#include <ATen/ATen.h>
+#include <ATen/cuda/CUDAContext.h>
+
+#include <THC/THCAtomics.cuh>
+
+#define CUDA_KERNEL_LOOP(i, n)                          \
+  for (int i = blockIdx.x * blockDim.x + threadIdx.x;   \
+      i < (n);                                          \
+      i += blockDim.x * gridDim.x)
+
+const int CUDA_NUM_THREADS = 1024;
+inline int GET_BLOCKS(const int N, const int num_threads)
+{
+  return (N + num_threads - 1) / num_threads;
+}
+
+
+template <typename scalar_t>
+__device__ scalar_t ms_deform_attn_im2col_bilinear(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+  }
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  return val;
+}
+
+
+template <typename scalar_t>
+__device__ void ms_deform_attn_col2im_bilinear(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c,
+                                                   const scalar_t &top_grad,
+                                                   const scalar_t &attn_weight,
+                                                   scalar_t* &grad_value, 
+                                                   scalar_t* grad_sampling_loc,
+                                                   scalar_t* grad_attn_weight)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+  const scalar_t top_grad_value = top_grad * attn_weight;
+  scalar_t grad_h_weight = 0, grad_w_weight = 0;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+    grad_h_weight -= hw * v1;
+    grad_w_weight -= hh * v1;
+    atomicAdd(grad_value+ptr1, w1*top_grad_value);
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+    grad_h_weight -= lw * v2;
+    grad_w_weight += hh * v2;
+    atomicAdd(grad_value+ptr2, w2*top_grad_value);
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+    grad_h_weight += hw * v3;
+    grad_w_weight -= lh * v3;
+    atomicAdd(grad_value+ptr3, w3*top_grad_value); 
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+    grad_h_weight += lw * v4;
+    grad_w_weight += lh * v4;
+    atomicAdd(grad_value+ptr4, w4*top_grad_value);
+  }
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  *grad_attn_weight = top_grad * val;
+  *grad_sampling_loc = width * grad_w_weight * top_grad_value;
+  *(grad_sampling_loc + 1) = height * grad_h_weight * top_grad_value;
+}
+
+
+template <typename scalar_t>
+__device__ void ms_deform_attn_col2im_bilinear_gm(const scalar_t* &bottom_data, 
+                                                   const int &height, const int &width, const int &nheads, const int &channels,
+                                                   const scalar_t &h, const scalar_t &w, const int &m, const int &c,
+                                                   const scalar_t &top_grad,
+                                                   const scalar_t &attn_weight,
+                                                   scalar_t* &grad_value, 
+                                                   scalar_t* grad_sampling_loc,
+                                                   scalar_t* grad_attn_weight)
+{
+  const int h_low = floor(h);
+  const int w_low = floor(w);
+  const int h_high = h_low + 1;
+  const int w_high = w_low + 1;
+
+  const scalar_t lh = h - h_low;
+  const scalar_t lw = w - w_low;
+  const scalar_t hh = 1 - lh, hw = 1 - lw;
+
+  const int w_stride = nheads * channels;
+  const int h_stride = width * w_stride;
+  const int h_low_ptr_offset = h_low * h_stride;
+  const int h_high_ptr_offset = h_low_ptr_offset + h_stride;
+  const int w_low_ptr_offset = w_low * w_stride;
+  const int w_high_ptr_offset = w_low_ptr_offset + w_stride;
+  const int base_ptr = m * channels + c;
+
+  const scalar_t w1 = hh * hw, w2 = hh * lw, w3 = lh * hw, w4 = lh * lw;
+  const scalar_t top_grad_value = top_grad * attn_weight;
+  scalar_t grad_h_weight = 0, grad_w_weight = 0;
+
+  scalar_t v1 = 0;
+  if (h_low >= 0 && w_low >= 0)
+  {
+    const int ptr1 = h_low_ptr_offset + w_low_ptr_offset + base_ptr;
+    v1 = bottom_data[ptr1];
+    grad_h_weight -= hw * v1;
+    grad_w_weight -= hh * v1;
+    atomicAdd(grad_value+ptr1, w1*top_grad_value);
+  }
+  scalar_t v2 = 0;
+  if (h_low >= 0 && w_high <= width - 1)
+  {
+    const int ptr2 = h_low_ptr_offset + w_high_ptr_offset + base_ptr;
+    v2 = bottom_data[ptr2];
+    grad_h_weight -= lw * v2;
+    grad_w_weight += hh * v2;
+    atomicAdd(grad_value+ptr2, w2*top_grad_value);
+  }
+  scalar_t v3 = 0;
+  if (h_high <= height - 1 && w_low >= 0)
+  {
+    const int ptr3 = h_high_ptr_offset + w_low_ptr_offset + base_ptr;
+    v3 = bottom_data[ptr3];
+    grad_h_weight += hw * v3;
+    grad_w_weight -= lh * v3;
+    atomicAdd(grad_value+ptr3, w3*top_grad_value); 
+  }
+  scalar_t v4 = 0;
+  if (h_high <= height - 1 && w_high <= width - 1)
+  {
+    const int ptr4 = h_high_ptr_offset + w_high_ptr_offset + base_ptr;
+    v4 = bottom_data[ptr4];
+    grad_h_weight += lw * v4;
+    grad_w_weight += lh * v4;
+    atomicAdd(grad_value+ptr4, w4*top_grad_value);
+  }
+
+  const scalar_t val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
+  atomicAdd(grad_attn_weight, top_grad * val); 
+  atomicAdd(grad_sampling_loc, width * grad_w_weight * top_grad_value);
+  atomicAdd(grad_sampling_loc + 1, height * grad_h_weight * top_grad_value);
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_im2col_gpu_kernel(const int n,
+                                                const scalar_t *data_value, 
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *data_col)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    scalar_t *data_col_ptr = data_col + index;
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+    scalar_t col = 0;
+    
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const scalar_t *data_value_ptr = data_value + (data_value_ptr_init_offset + level_start_id * qid_stride);
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          col += ms_deform_attn_im2col_bilinear(data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col) * weight;
+        }
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+      }
+    }
+    *data_col_ptr = col;
+  }
+}
+
+template <typename scalar_t, unsigned int blockSize>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2];
+    __shared__ scalar_t cache_grad_attn_weight[blockSize];
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+        if (tid == 0)
+        {
+          scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0];
+          int sid=2;
+          for (unsigned int tid = 1; tid < blockSize; ++tid)
+          {
+            _grad_w += cache_grad_sampling_loc[sid];
+            _grad_h += cache_grad_sampling_loc[sid + 1];
+            _grad_a += cache_grad_attn_weight[tid];
+            sid += 2;
+          }
+          
+          
+          *grad_sampling_loc = _grad_w;
+          *(grad_sampling_loc + 1) = _grad_h;
+          *grad_attn_weight = _grad_a;
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t, unsigned int blockSize>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    __shared__ scalar_t cache_grad_sampling_loc[blockSize * 2];
+    __shared__ scalar_t cache_grad_attn_weight[blockSize];
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockSize/2; s>0; s>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        { 
+          *grad_sampling_loc = cache_grad_sampling_loc[0];
+          *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1];
+          *grad_attn_weight = cache_grad_attn_weight[0];
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v1(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+        if (tid == 0)
+        {
+          scalar_t _grad_w=cache_grad_sampling_loc[0], _grad_h=cache_grad_sampling_loc[1], _grad_a=cache_grad_attn_weight[0];
+          int sid=2;
+          for (unsigned int tid = 1; tid < blockDim.x; ++tid)
+          {
+            _grad_w += cache_grad_sampling_loc[sid];
+            _grad_h += cache_grad_sampling_loc[sid + 1];
+            _grad_a += cache_grad_attn_weight[tid];
+            sid += 2;
+          }
+          
+          
+          *grad_sampling_loc = _grad_w;
+          *(grad_sampling_loc + 1) = _grad_h;
+          *grad_attn_weight = _grad_a;
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+            if (tid + (s << 1) < spre)
+            {
+              cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)];
+              cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)];
+              cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)];
+            } 
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        {
+          *grad_sampling_loc = cache_grad_sampling_loc[0];
+          *(grad_sampling_loc + 1) = cache_grad_sampling_loc[1];
+          *grad_attn_weight = cache_grad_attn_weight[0];
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    extern __shared__ int _s[];
+    scalar_t* cache_grad_sampling_loc = (scalar_t*)_s;
+    scalar_t* cache_grad_attn_weight = cache_grad_sampling_loc + 2 * blockDim.x;
+    unsigned int tid = threadIdx.x;
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        *(cache_grad_sampling_loc+(threadIdx.x << 1)) = 0;
+        *(cache_grad_sampling_loc+((threadIdx.x << 1) + 1)) = 0;
+        *(cache_grad_attn_weight+threadIdx.x)=0;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            cache_grad_sampling_loc+(threadIdx.x << 1), cache_grad_attn_weight+threadIdx.x);
+        }
+        
+        __syncthreads();
+
+        for (unsigned int s=blockDim.x/2, spre=blockDim.x; s>0; s>>=1, spre>>=1)
+        {
+          if (tid < s) {
+            const unsigned int xid1 = tid << 1;
+            const unsigned int xid2 = (tid + s) << 1;
+            cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + s];
+            cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2];
+            cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1];
+            if (tid + (s << 1) < spre)
+            {
+              cache_grad_attn_weight[tid] += cache_grad_attn_weight[tid + (s << 1)];
+              cache_grad_sampling_loc[xid1] += cache_grad_sampling_loc[xid2 + (s << 1)];
+              cache_grad_sampling_loc[xid1 + 1] += cache_grad_sampling_loc[xid2 + 1 + (s << 1)];
+            }
+          }
+          __syncthreads();
+        }
+
+        if (tid == 0)
+        {
+          atomicAdd(grad_sampling_loc, cache_grad_sampling_loc[0]);
+          atomicAdd(grad_sampling_loc + 1, cache_grad_sampling_loc[1]);
+          atomicAdd(grad_attn_weight, cache_grad_attn_weight[0]);
+        }
+        __syncthreads();
+
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+__global__ void ms_deformable_col2im_gpu_kernel_gm(const int n,
+                                                const scalar_t *grad_col,
+                                                const scalar_t *data_value,
+                                                const int64_t *data_spatial_shapes,
+                                                const int64_t *data_level_start_index, 
+                                                const scalar_t *data_sampling_loc,
+                                                const scalar_t *data_attn_weight,
+                                                const int batch_size, 
+                                                const int spatial_size, 
+                                                const int num_heads,
+                                                const int channels, 
+                                                const int num_levels,
+                                                const int num_query,
+                                                const int num_point,
+                                                scalar_t *grad_value,
+                                                scalar_t *grad_sampling_loc,
+                                                scalar_t *grad_attn_weight)
+{
+  CUDA_KERNEL_LOOP(index, n)
+  {
+    int _temp = index;
+    const int c_col = _temp % channels;
+    _temp /= channels;
+    const int sampling_index = _temp; 
+    const int m_col = _temp % num_heads;
+    _temp /= num_heads;
+    const int q_col = _temp % num_query;
+    _temp /= num_query;
+    const int b_col = _temp;
+
+    const scalar_t top_grad = grad_col[index];
+
+    int data_weight_ptr = sampling_index * num_levels * num_point;
+    int data_loc_w_ptr = data_weight_ptr << 1;
+    const int grad_sampling_ptr = data_weight_ptr;
+    grad_sampling_loc += grad_sampling_ptr << 1;
+    grad_attn_weight += grad_sampling_ptr;
+    const int grad_weight_stride = 1;
+    const int grad_loc_stride = 2;
+    const int qid_stride = num_heads * channels;
+    const int data_value_ptr_init_offset = b_col * spatial_size * qid_stride;
+
+    for (int l_col=0; l_col < num_levels; ++l_col)
+    {
+      const int level_start_id = data_level_start_index[l_col];
+      const int spatial_h_ptr = l_col << 1;
+      const int spatial_h = data_spatial_shapes[spatial_h_ptr];
+      const int spatial_w = data_spatial_shapes[spatial_h_ptr + 1];
+      const int value_ptr_offset = data_value_ptr_init_offset + level_start_id * qid_stride;
+      const scalar_t *data_value_ptr = data_value + value_ptr_offset;
+      scalar_t *grad_value_ptr = grad_value + value_ptr_offset;
+
+      for (int p_col=0; p_col < num_point; ++p_col)
+      {
+        const scalar_t loc_w = data_sampling_loc[data_loc_w_ptr];
+        const scalar_t loc_h = data_sampling_loc[data_loc_w_ptr + 1];
+        const scalar_t weight = data_attn_weight[data_weight_ptr];
+
+        const scalar_t h_im = loc_h * spatial_h - 0.5;
+        const scalar_t w_im = loc_w * spatial_w - 0.5;
+        if (h_im > -1 && w_im > -1 && h_im < spatial_h && w_im < spatial_w)
+        {
+          ms_deform_attn_col2im_bilinear_gm(
+            data_value_ptr, spatial_h, spatial_w, num_heads, channels, h_im, w_im, m_col, c_col,
+            top_grad, weight, grad_value_ptr, 
+            grad_sampling_loc, grad_attn_weight);
+        }
+        data_weight_ptr += 1;
+        data_loc_w_ptr += 2;
+        grad_attn_weight += grad_weight_stride;
+        grad_sampling_loc += grad_loc_stride;
+      }
+    }
+  }
+}
+
+
+template <typename scalar_t>
+void ms_deformable_im2col_cuda(cudaStream_t stream,
+                              const scalar_t* data_value,
+                              const int64_t* data_spatial_shapes, 
+                              const int64_t* data_level_start_index, 
+                              const scalar_t* data_sampling_loc,
+                              const scalar_t* data_attn_weight,
+                              const int batch_size,
+                              const int spatial_size, 
+                              const int num_heads, 
+                              const int channels, 
+                              const int num_levels, 
+                              const int num_query,
+                              const int num_point,
+                              scalar_t* data_col)
+{
+  const int num_kernels = batch_size * num_query * num_heads * channels;
+  const int num_actual_kernels = batch_size * num_query * num_heads * channels;
+  const int num_threads = CUDA_NUM_THREADS;
+  ms_deformable_im2col_gpu_kernel<scalar_t>
+      <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+          0, stream>>>(
+      num_kernels, data_value, data_spatial_shapes, data_level_start_index, data_sampling_loc, data_attn_weight, 
+      batch_size, spatial_size, num_heads, channels, num_levels, num_query, num_point, data_col);
+  
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess)
+  {
+    printf("error in ms_deformable_im2col_cuda: %s\n", cudaGetErrorString(err));
+  }
+
+}
+
+template <typename scalar_t>
+void ms_deformable_col2im_cuda(cudaStream_t stream,
+                              const scalar_t* grad_col,
+                              const scalar_t* data_value,
+                              const int64_t * data_spatial_shapes,
+                              const int64_t * data_level_start_index,
+                              const scalar_t * data_sampling_loc,
+                              const scalar_t * data_attn_weight,
+                              const int batch_size, 
+                              const int spatial_size, 
+                              const int num_heads,
+                              const int channels, 
+                              const int num_levels,
+                              const int num_query,
+                              const int num_point, 
+                              scalar_t* grad_value,
+                              scalar_t* grad_sampling_loc,
+                              scalar_t* grad_attn_weight)
+{
+  const int num_threads = (channels > CUDA_NUM_THREADS)?CUDA_NUM_THREADS:channels;
+  const int num_kernels = batch_size * num_query * num_heads * channels;
+  const int num_actual_kernels = batch_size * num_query * num_heads * channels;
+  if (channels > 1024)
+  {
+    if ((channels & 1023) == 0)
+    {
+      ms_deformable_col2im_gpu_kernel_shm_reduce_v2_multi_blocks<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+    }
+    else
+    {
+      ms_deformable_col2im_gpu_kernel_gm<scalar_t>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+    }
+  }
+  else{
+    switch(channels)
+    {
+      case 1:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 1>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 2:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 2>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 4:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 4>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 8:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 8>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 16:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 16>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 32:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v1<scalar_t, 32>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 64:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 64>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 128:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 128>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 256:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 256>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 512:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 512>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      case 1024:
+        ms_deformable_col2im_gpu_kernel_shm_blocksize_aware_reduce_v2<scalar_t, 1024>
+        <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+            0, stream>>>(
+                      num_kernels, 
+                      grad_col,
+                      data_value,
+                      data_spatial_shapes,
+                      data_level_start_index, 
+                      data_sampling_loc,
+                      data_attn_weight,
+                      batch_size, 
+                      spatial_size, 
+                      num_heads,
+                      channels, 
+                      num_levels,
+                      num_query,
+                      num_point,
+                      grad_value,
+                      grad_sampling_loc,
+                      grad_attn_weight);
+        break;
+      default:
+        if (channels < 64)
+        {
+          ms_deformable_col2im_gpu_kernel_shm_reduce_v1<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+        }
+        else
+        {
+          ms_deformable_col2im_gpu_kernel_shm_reduce_v2<scalar_t>
+          <<<GET_BLOCKS(num_actual_kernels, num_threads), num_threads,
+              num_threads*3*sizeof(scalar_t), stream>>>(
+                        num_kernels, 
+                        grad_col,
+                        data_value,
+                        data_spatial_shapes,
+                        data_level_start_index, 
+                        data_sampling_loc,
+                        data_attn_weight,
+                        batch_size, 
+                        spatial_size, 
+                        num_heads,
+                        channels, 
+                        num_levels,
+                        num_query,
+                        num_point,
+                        grad_value,
+                        grad_sampling_loc,
+                        grad_attn_weight);
+        }
+    }
+  }
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess)
+  {
+    printf("error in ms_deformable_col2im_cuda: %s\n", cudaGetErrorString(err));
+  }
+
+}

phivenv/Lib/site-packages/transformers/kernels/deta/ms_deform_attn.h

@@ -0,0 +1,61 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#pragma once
+
+#include "cpu/ms_deform_attn_cpu.h"
+
+#ifdef WITH_CUDA
+#include "cuda/ms_deform_attn_cuda.h"
+#endif
+
+
+at::Tensor
+ms_deform_attn_forward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const int im2col_step)
+{
+    if (value.type().is_cuda())
+    {
+#ifdef WITH_CUDA
+        return ms_deform_attn_cuda_forward(
+            value, spatial_shapes, level_start_index, sampling_loc, attn_weight, im2col_step);
+#else
+        AT_ERROR("Not compiled with GPU support");
+#endif
+    }
+    AT_ERROR("Not implemented on the CPU");
+}
+
+std::vector<at::Tensor>
+ms_deform_attn_backward(
+    const at::Tensor &value, 
+    const at::Tensor &spatial_shapes,
+    const at::Tensor &level_start_index,
+    const at::Tensor &sampling_loc,
+    const at::Tensor &attn_weight,
+    const at::Tensor &grad_output,
+    const int im2col_step)
+{
+    if (value.type().is_cuda())
+    {
+#ifdef WITH_CUDA
+        return ms_deform_attn_cuda_backward(
+            value, spatial_shapes, level_start_index, sampling_loc, attn_weight, grad_output, im2col_step);
+#else
+        AT_ERROR("Not compiled with GPU support");
+#endif
+    }
+    AT_ERROR("Not implemented on the CPU");
+}

phivenv/Lib/site-packages/transformers/kernels/deta/vision.cpp

@@ -0,0 +1,16 @@
+/*!
+**************************************************************************************************
+* Deformable DETR
+* Copyright (c) 2020 SenseTime. All Rights Reserved.
+* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+**************************************************************************************************
+* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
+**************************************************************************************************
+*/
+
+#include "ms_deform_attn.h"
+
+PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
+  m.def("ms_deform_attn_forward", &ms_deform_attn_forward, "ms_deform_attn_forward");
+  m.def("ms_deform_attn_backward", &ms_deform_attn_backward, "ms_deform_attn_backward");
+}

phivenv/Lib/site-packages/transformers/kernels/falcon_mamba/__init__.py

@@ -0,0 +1,15 @@
+# coding=utf-8
+# Copyright 2024 Tri Dao, Albert Gu, Technological Innovation Institute and HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from .selective_scan_with_ln_interface import mamba_inner_fn

phivenv/Lib/site-packages/transformers/kernels/falcon_mamba/selective_scan_with_ln_interface.py

@@ -0,0 +1,525 @@
+# coding=utf-8
+# Copyright 2024 Tri Dao, Albert Gu, Technological Innovation Institute and HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# Original code from: https://github.com/state-spaces/mamba/blob/main/mamba_ssm/ops/selective_scan_interface.py
+
+import torch
+import torch.nn.functional as F
+from einops import rearrange, repeat
+from torch.cuda.amp import custom_bwd, custom_fwd
+
+
+try:
+    import causal_conv1d_cuda
+except ImportError:
+    causal_conv1d_cuda = None
+
+import mamba_ssm
+import selective_scan_cuda
+
+
+# For BC for old mamba-ssm versions: https://github.com/huggingface/transformers/pull/33195#discussion_r1736401127
+if hasattr(mamba_ssm.ops.triton, "layernorm"):
+    from mamba_ssm.ops.triton.layernorm import _layer_norm_fwd
+else:
+    from mamba_ssm.ops.triton.layer_norm import _layer_norm_fwd
+
+
+class SelectiveScanFn(torch.autograd.Function):
+    @staticmethod
+    def forward(
+        ctx, u, delta, A, B, C, D=None, z=None, delta_bias=None, delta_softplus=False, return_last_state=False
+    ):
+        if u.stride(-1) != 1:
+            u = u.contiguous()
+        if delta.stride(-1) != 1:
+            delta = delta.contiguous()
+        if D is not None:
+            D = D.contiguous()
+        if B.stride(-1) != 1:
+            B = B.contiguous()
+        if C.stride(-1) != 1:
+            C = C.contiguous()
+        if z is not None and z.stride(-1) != 1:
+            z = z.contiguous()
+        if B.dim() == 3:
+            B = rearrange(B, "b dstate l -> b 1 dstate l")
+            ctx.squeeze_B = True
+        if C.dim() == 3:
+            C = rearrange(C, "b dstate l -> b 1 dstate l")
+            ctx.squeeze_C = True
+        out, x, *rest = selective_scan_cuda.fwd(u, delta, A, B, C, D, z, delta_bias, delta_softplus)
+        ctx.delta_softplus = delta_softplus
+        ctx.has_z = z is not None
+        last_state = x[:, :, -1, 1::2]  # (batch, dim, dstate)
+        if not ctx.has_z:
+            ctx.save_for_backward(u, delta, A, B, C, D, delta_bias, x)
+            return out if not return_last_state else (out, last_state)
+        else:
+            ctx.save_for_backward(u, delta, A, B, C, D, z, delta_bias, x, out)
+            out_z = rest[0]
+            return out_z if not return_last_state else (out_z, last_state)
+
+    @staticmethod
+    def backward(ctx, dout, *args):
+        if not ctx.has_z:
+            u, delta, A, B, C, D, delta_bias, x = ctx.saved_tensors
+            z = None
+            out = None
+        else:
+            u, delta, A, B, C, D, z, delta_bias, x, out = ctx.saved_tensors
+        if dout.stride(-1) != 1:
+            dout = dout.contiguous()
+        # The kernel supports passing in a pre-allocated dz (e.g., in case we want to fuse the
+        # backward of selective_scan_cuda with the backward of chunk).
+        # Here we just pass in None and dz will be allocated in the C++ code.
+        du, ddelta, dA, dB, dC, dD, ddelta_bias, *rest = selective_scan_cuda.bwd(
+            u,
+            delta,
+            A,
+            B,
+            C,
+            D,
+            z,
+            delta_bias,
+            dout,
+            x,
+            out,
+            None,
+            ctx.delta_softplus,
+            False,  # option to recompute out_z, not used here
+        )
+        dz = rest[0] if ctx.has_z else None
+        dB = dB.squeeze(1) if getattr(ctx, "squeeze_B", False) else dB
+        dC = dC.squeeze(1) if getattr(ctx, "squeeze_C", False) else dC
+        return (
+            du,
+            ddelta,
+            dA,
+            dB,
+            dC,
+            dD if D is not None else None,
+            dz,
+            ddelta_bias if delta_bias is not None else None,
+            None,
+            None,
+        )
+
+
+def rms_norm_forward(
+    x,
+    weight,
+    bias,
+    eps=1e-6,
+    is_rms_norm=True,
+):
+    # x (b l) d
+    if x.stride(-1) != 1:
+        x = x.contiguous()
+    weight = weight.contiguous()
+    if bias is not None:
+        bias = bias.contiguous()
+    y = _layer_norm_fwd(x, weight, bias, eps, None, residual_dtype=None, is_rms_norm=is_rms_norm)[0]
+    # y (b l) d
+    return y
+
+
+def selective_scan_fn(
+    u, delta, A, B, C, D=None, z=None, delta_bias=None, delta_softplus=False, return_last_state=False
+):
+    """if return_last_state is True, returns (out, last_state)
+    last_state has shape (batch, dim, dstate). Note that the gradient of the last state is
+    not considered in the backward pass.
+    """
+    return SelectiveScanFn.apply(u, delta, A, B, C, D, z, delta_bias, delta_softplus, return_last_state)
+
+
+def selective_scan_ref(
+    u, delta, A, B, C, D=None, z=None, delta_bias=None, delta_softplus=False, return_last_state=False
+):
+    """
+    u: r(B D L)
+    delta: r(B D L)
+    A: c(D N) or r(D N)
+    B: c(D N) or r(B N L) or r(B N 2L) or r(B G N L) or (B G N L)
+    C: c(D N) or r(B N L) or r(B N 2L) or r(B G N L) or (B G N L)
+    D: r(D)
+    z: r(B D L)
+    delta_bias: r(D), fp32
+
+    out: r(B D L)
+    last_state (optional): r(B D dstate) or c(B D dstate)
+    """
+    dtype_in = u.dtype
+    u = u.float()
+    delta = delta.float()
+    if delta_bias is not None:
+        delta = delta + delta_bias[..., None].float()
+    if delta_softplus:
+        delta = F.softplus(delta)
+    batch, dim, dstate = u.shape[0], A.shape[0], A.shape[1]
+    is_variable_B = B.dim() >= 3
+    is_variable_C = C.dim() >= 3
+    if A.is_complex():
+        if is_variable_B:
+            B = torch.view_as_complex(rearrange(B.float(), "... (L two) -> ... L two", two=2))
+        if is_variable_C:
+            C = torch.view_as_complex(rearrange(C.float(), "... (L two) -> ... L two", two=2))
+    else:
+        B = B.float()
+        C = C.float()
+    x = A.new_zeros((batch, dim, dstate))
+    ys = []
+    deltaA = torch.exp(torch.einsum("bdl,dn->bdln", delta, A))
+    if not is_variable_B:
+        deltaB_u = torch.einsum("bdl,dn,bdl->bdln", delta, B, u)
+    else:
+        if B.dim() == 3:
+            deltaB_u = torch.einsum("bdl,bnl,bdl->bdln", delta, B, u)
+        else:
+            B = repeat(B, "B G N L -> B (G H) N L", H=dim // B.shape[1])
+            deltaB_u = torch.einsum("bdl,bdnl,bdl->bdln", delta, B, u)
+    if is_variable_C and C.dim() == 4:
+        C = repeat(C, "B G N L -> B (G H) N L", H=dim // C.shape[1])
+    last_state = None
+    for i in range(u.shape[2]):
+        x = deltaA[:, :, i] * x + deltaB_u[:, :, i]
+        if not is_variable_C:
+            y = torch.einsum("bdn,dn->bd", x, C)
+        else:
+            if C.dim() == 3:
+                y = torch.einsum("bdn,bn->bd", x, C[:, :, i])
+            else:
+                y = torch.einsum("bdn,bdn->bd", x, C[:, :, :, i])
+        if i == u.shape[2] - 1:
+            last_state = x
+        if y.is_complex():
+            y = y.real * 2
+        ys.append(y)
+    y = torch.stack(ys, dim=2)  # (batch dim L)
+    out = y if D is None else y + u * rearrange(D, "d -> d 1")
+    if z is not None:
+        out = out * F.silu(z)
+    out = out.to(dtype=dtype_in)
+    return out if not return_last_state else (out, last_state)
+
+
+class MambaInnerFn(torch.autograd.Function):
+    @staticmethod
+    @custom_fwd
+    def forward(
+        ctx,
+        xz,
+        conv1d_weight,
+        conv1d_bias,
+        x_proj_weight,
+        delta_proj_weight,
+        out_proj_weight,
+        out_proj_bias,
+        A,
+        B=None,
+        C=None,
+        D=None,
+        delta_bias=None,
+        B_proj_bias=None,
+        C_proj_bias=None,
+        delta_softplus=True,
+        checkpoint_lvl=1,
+        b_rms_weight=None,
+        c_rms_weight=None,
+        dt_rms_weight=None,
+        b_c_dt_rms_eps=1e-6,
+    ):
+        """
+        xz: (batch, dim, seqlen)
+        """
+        assert causal_conv1d_cuda is not None, "causal_conv1d_cuda is not available. Please install causal-conv1d."
+        assert checkpoint_lvl in [0, 1]
+        L = xz.shape[-1]
+        delta_rank = delta_proj_weight.shape[1]
+        d_state = A.shape[-1] * (1 if not A.is_complex() else 2)
+        if torch.is_autocast_enabled():
+            x_proj_weight = x_proj_weight.to(dtype=torch.get_autocast_gpu_dtype())
+            delta_proj_weight = delta_proj_weight.to(dtype=torch.get_autocast_gpu_dtype())
+            out_proj_weight = out_proj_weight.to(dtype=torch.get_autocast_gpu_dtype())
+            out_proj_bias = (
+                out_proj_bias.to(dtype=torch.get_autocast_gpu_dtype()) if out_proj_bias is not None else None
+            )
+        if xz.stride(-1) != 1:
+            xz = xz.contiguous()
+        conv1d_weight = rearrange(conv1d_weight, "d 1 w -> d w")
+        x, z = xz.chunk(2, dim=1)
+        conv1d_bias = conv1d_bias.contiguous() if conv1d_bias is not None else None
+        conv1d_out = causal_conv1d_cuda.causal_conv1d_fwd(x, conv1d_weight, conv1d_bias, None, None, None, True)
+        # We're being very careful here about the layout, to avoid extra transposes.
+        # We want delta to have d as the slowest moving dimension
+        # and L as the fastest moving dimension, since those are what the ssm_scan kernel expects.
+        x_dbl = F.linear(rearrange(conv1d_out, "b d l -> (b l) d"), x_proj_weight)  # (bl d)
+        delta = rearrange(delta_proj_weight @ x_dbl[:, :delta_rank].t(), "d (b l) -> b d l", l=L)
+        ctx.is_variable_B = B is None
+        ctx.is_variable_C = C is None
+        ctx.B_proj_bias_is_None = B_proj_bias is None
+        ctx.C_proj_bias_is_None = C_proj_bias is None
+        if B is None:  # variable B
+            B = x_dbl[:, delta_rank : delta_rank + d_state]  # (bl dstate)
+            if B_proj_bias is not None:
+                B = B + B_proj_bias.to(dtype=B.dtype)
+            if not A.is_complex():
+                # B = rearrange(B, "(b l) dstate -> b dstate l", l=L).contiguous()
+                B = rearrange(B, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+            else:
+                B = rearrange(B, "(b l) (dstate two) -> b 1 dstate (l two)", l=L, two=2).contiguous()
+        else:
+            if B.stride(-1) != 1:
+                B = B.contiguous()
+        if C is None:  # variable C
+            C = x_dbl[:, -d_state:]  # (bl dstate)
+            if C_proj_bias is not None:
+                C = C + C_proj_bias.to(dtype=C.dtype)
+            if not A.is_complex():
+                # C = rearrange(C, "(b l) dstate -> b dstate l", l=L).contiguous()
+                C = rearrange(C, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+            else:
+                C = rearrange(C, "(b l) (dstate two) -> b 1 dstate (l two)", l=L, two=2).contiguous()
+        else:
+            if C.stride(-1) != 1:
+                C = C.contiguous()
+        if D is not None:
+            D = D.contiguous()
+
+        if b_rms_weight is not None:
+            B = rearrange(B, "b 1 dstate l -> (b l) dstate", l=L).contiguous()
+            B = rms_norm_forward(B, b_rms_weight, bias=None, eps=b_c_dt_rms_eps)
+            B = rearrange(B, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+        if c_rms_weight is not None:
+            C = rearrange(C, "b 1 dstate l -> (b l) dstate", l=L).contiguous()
+            C = rms_norm_forward(C, c_rms_weight, bias=None, eps=b_c_dt_rms_eps)
+            C = rearrange(C, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+        if dt_rms_weight is not None:
+            delta = rearrange(delta, "b d l -> (b l) d", l=L).contiguous()
+            delta = rms_norm_forward(delta, dt_rms_weight, bias=None, eps=b_c_dt_rms_eps)
+            delta = rearrange(delta, "(b l) d -> b d l", l=L).contiguous()
+
+        out, scan_intermediates, out_z = selective_scan_cuda.fwd(
+            conv1d_out, delta, A, B, C, D, z, delta_bias, delta_softplus
+        )
+        ctx.delta_softplus = delta_softplus
+        ctx.out_proj_bias_is_None = out_proj_bias is None
+        ctx.checkpoint_lvl = checkpoint_lvl
+        ctx.b_rms_weight = b_rms_weight
+        ctx.c_rms_weight = c_rms_weight
+        ctx.dt_rms_weight = dt_rms_weight
+        ctx.b_c_dt_rms_eps = b_c_dt_rms_eps
+        if checkpoint_lvl >= 1:  # Will recompute conv1d_out and delta in the backward pass
+            conv1d_out, delta = None, None
+        ctx.save_for_backward(
+            xz,
+            conv1d_weight,
+            conv1d_bias,
+            x_dbl,
+            x_proj_weight,
+            delta_proj_weight,
+            out_proj_weight,
+            conv1d_out,
+            delta,
+            A,
+            B,
+            C,
+            D,
+            delta_bias,
+            scan_intermediates,
+            b_rms_weight,
+            c_rms_weight,
+            dt_rms_weight,
+            out,
+        )
+        return F.linear(rearrange(out_z, "b d l -> b l d"), out_proj_weight, out_proj_bias)
+
+    @staticmethod
+    @custom_bwd
+    def backward(ctx, dout):
+        # dout: (batch, seqlen, dim)
+        assert causal_conv1d_cuda is not None, "causal_conv1d_cuda is not available. Please install causal-conv1d."
+        (
+            xz,
+            conv1d_weight,
+            conv1d_bias,
+            x_dbl,
+            x_proj_weight,
+            delta_proj_weight,
+            out_proj_weight,
+            conv1d_out,
+            delta,
+            A,
+            B,
+            C,
+            D,
+            delta_bias,
+            scan_intermediates,
+            b_rms_weight,
+            c_rms_weight,
+            dt_rms_weight,
+            out,
+        ) = ctx.saved_tensors
+        L = xz.shape[-1]
+        delta_rank = delta_proj_weight.shape[1]
+        d_state = A.shape[-1] * (1 if not A.is_complex() else 2)
+        x, z = xz.chunk(2, dim=1)
+        if dout.stride(-1) != 1:
+            dout = dout.contiguous()
+        if ctx.checkpoint_lvl == 1:
+            conv1d_out = causal_conv1d_cuda.causal_conv1d_fwd(x, conv1d_weight, conv1d_bias, None, None, None, True)
+            delta = rearrange(delta_proj_weight @ x_dbl[:, :delta_rank].t(), "d (b l) -> b d l", l=L)
+            if dt_rms_weight is not None:
+                delta = rearrange(delta, "b d l -> (b l) d", l=L).contiguous()
+                delta = rms_norm_forward(delta, ctx.dt_rms_weight, None, ctx.b_c_dt_rms_eps)
+                delta = rearrange(delta, "(b l) d -> b d l", l=L).contiguous()
+            if b_rms_weight is not None:
+                # Recompute & RMSNorm B
+                B = rearrange(B, "b 1 dstate l -> (b l) dstate", l=L).contiguous()
+                B = rms_norm_forward(B, ctx.b_rms_weight, None, ctx.b_c_dt_rms_eps)
+                B = rearrange(B, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+            if c_rms_weight is not None:
+                # Recompute & RMSNorm C
+                C = rearrange(C, "b 1 dstate l -> (b l) dstate", l=L).contiguous()
+                C = rms_norm_forward(C, ctx.c_rms_weight, None, ctx.b_c_dt_rms_eps)
+                C = rearrange(C, "(b l) dstate -> b 1 dstate l", l=L).contiguous()
+
+        # The kernel supports passing in a pre-allocated dz (e.g., in case we want to fuse the
+        # backward of selective_scan_cuda with the backward of chunk).
+        dxz = torch.empty_like(xz)  # (batch, dim, seqlen)
+        dx, dz = dxz.chunk(2, dim=1)
+        dout = rearrange(dout, "b l e -> e (b l)")
+        dout_y = rearrange(out_proj_weight.t() @ dout, "d (b l) -> b d l", l=L)
+        dconv1d_out, ddelta, dA, dB, dC, dD, ddelta_bias, dz, out_z = selective_scan_cuda.bwd(
+            conv1d_out,
+            delta,
+            A,
+            B,
+            C,
+            D,
+            z,
+            delta_bias,
+            dout_y,
+            scan_intermediates,
+            out,
+            dz,
+            ctx.delta_softplus,
+            True,  # option to recompute out_z
+        )
+        dout_proj_weight = torch.einsum("eB,dB->ed", dout, rearrange(out_z, "b d l -> d (b l)"))
+        dout_proj_bias = dout.sum(dim=(0, 1)) if not ctx.out_proj_bias_is_None else None
+        dD = dD if D is not None else None
+        dx_dbl = torch.empty_like(x_dbl)
+        dB_proj_bias = None
+        if ctx.is_variable_B:
+            if not A.is_complex():
+                dB = rearrange(dB, "b 1 dstate l -> (b l) dstate").contiguous()
+            else:
+                dB = rearrange(dB, "b 1 dstate (l two) -> (b l) (dstate two)", two=2).contiguous()
+            dB_proj_bias = dB.sum(0) if not ctx.B_proj_bias_is_None else None
+            dx_dbl[:, delta_rank : delta_rank + d_state] = dB  # (bl d)
+            dB = None
+        dC_proj_bias = None
+        if ctx.is_variable_C:
+            if not A.is_complex():
+                dC = rearrange(dC, "b 1 dstate l -> (b l) dstate").contiguous()
+            else:
+                dC = rearrange(dC, "b 1 dstate (l two) -> (b l) (dstate two)", two=2).contiguous()
+            dC_proj_bias = dC.sum(0) if not ctx.C_proj_bias_is_None else None
+            dx_dbl[:, -d_state:] = dC  # (bl d)
+            dC = None
+        ddelta = rearrange(ddelta, "b d l -> d (b l)")
+        ddelta_proj_weight = torch.einsum("dB,Br->dr", ddelta, x_dbl[:, :delta_rank])
+        dx_dbl[:, :delta_rank] = torch.einsum("dB,dr->Br", ddelta, delta_proj_weight)
+        dconv1d_out = rearrange(dconv1d_out, "b d l -> d (b l)")
+        dx_proj_weight = torch.einsum("Br,Bd->rd", dx_dbl, rearrange(conv1d_out, "b d l -> (b l) d"))
+        dconv1d_out = torch.addmm(dconv1d_out, x_proj_weight.t(), dx_dbl.t(), out=dconv1d_out)
+        dconv1d_out = rearrange(dconv1d_out, "d (b l) -> b d l", b=x.shape[0], l=x.shape[-1])
+        # The kernel supports passing in a pre-allocated dx (e.g., in case we want to fuse the
+        # backward of conv1d with the backward of chunk).
+        dx, dconv1d_weight, dconv1d_bias, *_ = causal_conv1d_cuda.causal_conv1d_bwd(
+            x, conv1d_weight, conv1d_bias, dconv1d_out, None, None, None, dx, False, True
+        )
+        dconv1d_bias = dconv1d_bias if conv1d_bias is not None else None
+        dconv1d_weight = rearrange(dconv1d_weight, "d w -> d 1 w")
+        return (
+            dxz,
+            dconv1d_weight,
+            dconv1d_bias,
+            dx_proj_weight,
+            ddelta_proj_weight,
+            dout_proj_weight,
+            dout_proj_bias,
+            dA,
+            dB,
+            dC,
+            dD,
+            ddelta_bias if delta_bias is not None else None,
+            # 6-None are delta_softplus, checkpoint_lvl, b_rms_weight, c_rms_weight, dt_rms_weight, b_c_dt_rms_eps
+            dB_proj_bias,
+            dC_proj_bias,
+            None,
+            None,
+            None,
+            None,
+            None,
+            None,
+        )
+
+
+def mamba_inner_fn(
+    xz,
+    conv1d_weight,
+    conv1d_bias,
+    x_proj_weight,
+    delta_proj_weight,
+    out_proj_weight,
+    out_proj_bias,
+    A,
+    B=None,
+    C=None,
+    D=None,
+    delta_bias=None,
+    B_proj_bias=None,
+    C_proj_bias=None,
+    delta_softplus=True,
+    checkpoint_lvl=1,
+    b_rms_weight=None,
+    c_rms_weight=None,
+    dt_rms_weight=None,
+    b_c_dt_rms_eps=1e-6,
+):
+    return MambaInnerFn.apply(
+        xz,
+        conv1d_weight,
+        conv1d_bias,
+        x_proj_weight,
+        delta_proj_weight,
+        out_proj_weight,
+        out_proj_bias,
+        A,
+        B,
+        C,
+        D,
+        delta_bias,
+        B_proj_bias,
+        C_proj_bias,
+        delta_softplus,
+        checkpoint_lvl,
+        b_rms_weight,
+        c_rms_weight,
+        dt_rms_weight,
+        b_c_dt_rms_eps,
+    )

phivenv/Lib/site-packages/transformers/kernels/mra/cuda_kernel.cu

@@ -0,0 +1,383 @@
+#include "cuda_kernel.h"
+
+//////////////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////
+
+__global__ void index_max_cuda_kernel(
+  float *index_vals,       // [batch_size, 32, num_block]
+  int   *indices,        // [batch_size, num_block]
+  float *max_vals,        // [batch_size, A_num_block * 32]
+  float *max_vals_scatter,   // [batch_size, 32, num_block]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+) {
+
+  long batch_idx = blockIdx.x;
+
+  long thread_idx = threadIdx.x;
+  long num_thread = blockDim.x;
+
+  extern __shared__ float buffer[];
+  int *max_buffer = (int*)buffer;
+
+  for (int i = 0; i < A_num_block * 32; i = i + num_thread) {
+    int idx = i + thread_idx;
+    if (idx < A_num_block * 32) {
+      max_buffer[idx] = -1e8;
+    }
+  }
+  __syncthreads();
+
+  int *indices_pt = &indices[batch_idx * num_block];
+  float *index_vals_pt = &index_vals[batch_idx * num_block * 32];
+
+  for (int idx_start = 0; idx_start < 32 * num_block; idx_start = idx_start + num_thread) {
+    int idx = idx_start + thread_idx;
+    int A_block_idx = indices_pt[idx % num_block] / B_num_block;
+    atomicMax(&max_buffer[A_block_idx * 32 + idx / num_block], (int)(index_vals_pt[idx] * 1000));
+  }
+  __syncthreads();
+  
+  float *max_vals_pt = &max_vals[batch_idx * A_num_block * 32];
+  for (int i = 0; i < A_num_block * 32; i = i + num_thread) {
+    int idx = i + thread_idx;
+    if (idx < A_num_block * 32) {
+      max_vals_pt[idx] = (float)max_buffer[idx] / 1000.;
+    }
+  }
+  
+  float *max_vals_scatter_pt = &max_vals_scatter[batch_idx * num_block * 32];
+  for (int idx_start = 0; idx_start < 32 * num_block; idx_start = idx_start + num_thread) {
+    int idx = idx_start + thread_idx;
+    int A_block_idx = indices_pt[idx % num_block] / B_num_block;
+    max_vals_scatter_pt[idx] = (float)max_buffer[A_block_idx * 32 + idx / num_block] / 1000.;
+  }
+
+}
+
+__global__ void mm_to_sparse_cuda_kernel(
+  float *dense_A,   // [batch_size, A_num_block, dim, 32]
+  float *dense_B,   // [batch_size, B_num_block, dim, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *sparse_C,  // [batch_size, num_block, 32, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long dim,
+  long num_block
+) {
+
+  long batch_idx = blockIdx.y;
+  long block_idx = blockIdx.x * blockDim.y + threadIdx.y;
+
+  long thread_idx = threadIdx.x;
+
+  __shared__ float buffer[4096];
+  float *A_buffer = &buffer[threadIdx.y * 1024]; // [2, 8, 32]
+  float *B_buffer = &buffer[threadIdx.y * 1024 + 512]; // [2, 8, 32]
+
+  long batch_idx__block_idx = batch_idx * num_block + block_idx;
+
+  long AB_block_idx = indices[batch_idx__block_idx];
+  float *dense_A_pt = &dense_A[(batch_idx * A_num_block + AB_block_idx / B_num_block) * dim * 32];
+  float *dense_B_pt = &dense_B[(batch_idx * B_num_block + AB_block_idx % B_num_block) * dim * 32];
+
+  int reg_1_idx = thread_idx / 8;    // [0000000011111111222222223333333344444444555555556666666677777777]
+  int reg_2_idx = thread_idx % 8;    // [0123456701234567012345670123456701234567012345670123456701234567]
+
+  float reg_1[8];
+  float reg_2[8];
+
+  float reg_array[16] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+  #pragma unroll
+  for (int i = 0; i < 4; i++) {
+    A_buffer[i * 64 + thread_idx] = dense_A_pt[i * 64 + thread_idx];
+    B_buffer[i * 64 + thread_idx] = dense_B_pt[i * 64 + thread_idx];
+  }
+
+  __syncthreads();
+
+  #pragma unroll
+  for (int i = 0; i < 4; i++) {
+    reg_1[i] = A_buffer[reg_1_idx * 4 + i];
+    reg_2[i] = B_buffer[reg_2_idx * 4 + i];
+  }
+
+  for (int dim_stride = 1; dim_stride < (dim / 8); dim_stride++) {
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      A_buffer[(dim_stride % 2) * 256 + i * 64 + thread_idx] = dense_A_pt[dim_stride * 256 + i * 64 + thread_idx];
+      B_buffer[(dim_stride % 2) * 256 + i * 64 + thread_idx] = dense_B_pt[dim_stride * 256 + i * 64 + thread_idx];
+    }
+
+    #pragma unroll
+    for (int mini_dim_idx = 1; mini_dim_idx < 8; mini_dim_idx++) {
+      #pragma unroll
+      for (int i = 0; i < 4; i++) {
+        reg_1[(mini_dim_idx % 2) * 4 + i] = A_buffer[((dim_stride - 1) % 2) * 256 + mini_dim_idx * 32 + reg_1_idx * 4 + i];
+        reg_2[(mini_dim_idx % 2) * 4 + i] = B_buffer[((dim_stride - 1) % 2) * 256 + mini_dim_idx * 32 + reg_2_idx * 4 + i];
+      }
+      #pragma unroll
+      for (int i = 0; i < 4; i++) {
+        #pragma unroll
+        for (int j = 0; j < 4; j++) {
+          reg_array[i * 4 + j] += reg_1[((mini_dim_idx - 1) % 2) * 4 + i] * reg_2[((mini_dim_idx - 1) % 2) * 4 + j];
+        }
+      }
+    }
+
+    __syncthreads();
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      reg_1[i] = A_buffer[(dim_stride % 2) * 256 + reg_1_idx * 4 + i];
+      reg_2[i] = B_buffer[(dim_stride % 2) * 256 + reg_2_idx * 4 + i];
+    }
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      #pragma unroll
+      for (int j = 0; j < 4; j++) {
+        reg_array[i * 4 + j] += reg_1[4 + i] * reg_2[4 + j];
+      }
+    }
+
+  }
+
+  #pragma unroll
+  for (int mini_dim_idx = 1; mini_dim_idx < 8; mini_dim_idx++) {
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      reg_1[(mini_dim_idx % 2) * 4 + i] = A_buffer[256 + mini_dim_idx * 32 + reg_1_idx * 4 + i];
+      reg_2[(mini_dim_idx % 2) * 4 + i] = B_buffer[256 + mini_dim_idx * 32 + reg_2_idx * 4 + i];
+    }
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      #pragma unroll
+      for (int j = 0; j < 4; j++) {
+        reg_array[i * 4 + j] += reg_1[((mini_dim_idx - 1) % 2) * 4 + i] * reg_2[((mini_dim_idx - 1) % 2) * 4 + j];
+      }
+    }
+  }
+  #pragma unroll
+  for (int i = 0; i < 4; i++) {
+    #pragma unroll
+    for (int j = 0; j < 4; j++) {
+      reg_array[i * 4 + j] += reg_1[4 + i] * reg_2[4 + j];
+    }
+  }
+  __syncthreads();
+
+  float *C_buffer = &buffer[threadIdx.y * 1024]; // [32, 32]
+
+  #pragma unroll
+  for (int i = 0; i < 4; i++) {
+    #pragma unroll
+    for (int j = 0; j < 4; j++) {
+      C_buffer[(reg_2_idx * 4 + j) * 32 + reg_1_idx * 4 + i] = reg_array[i * 4 + j];
+    }
+  }
+  __syncthreads();
+
+  float *sparse_C_pt = &sparse_C[batch_idx__block_idx * 1024];
+
+  #pragma unroll
+  for (int i = 0; i < 16; i++) {
+    sparse_C_pt[i * 64 + thread_idx] = C_buffer[i * 64 + thread_idx];
+  }
+
+}
+
+__global__ void sparse_dense_mm_cuda_kernel(
+  float *sparse_A,  // [batch_size, num_block, 32, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *dense_B,   // [batch_size, B_num_block, dim, 32]
+  float *dense_C,   // [batch_size, A_num_block, dim, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long dim,
+  long num_block
+) {
+
+  long batch_idx = blockIdx.y;
+  long block_idx = blockIdx.x * blockDim.y + threadIdx.y;
+
+  long thread_idx = threadIdx.x;
+
+  __shared__ float buffer[6144];
+  float *A_buffer = &buffer[threadIdx.y * 3072]; // [32, 32]
+  float *B_buffer = &buffer[threadIdx.y * 3072 + 1024]; // [32, 64]
+
+  long batch_idx__block_idx = batch_idx * num_block + block_idx;
+
+  float *sparse_A_pt = &sparse_A[batch_idx__block_idx * 1024];
+  #pragma unroll
+  for (int i = 0; i < 8; i++) {
+    A_buffer[i * 128 + thread_idx] = sparse_A_pt[i * 128 + thread_idx];
+  }
+
+  long AB_block_idx = indices[batch_idx__block_idx];
+  float *dense_B_pt = &dense_B[(batch_idx * B_num_block + AB_block_idx % B_num_block) * 32 * dim];
+  float *dense_C_pt = &dense_C[(batch_idx * A_num_block + AB_block_idx / B_num_block) * 32 * dim];
+
+  // [0000000011111111222222223333333344444444555555556666666677777777]
+  // [0123456701234567012345670123456701234567012345670123456701234567]
+  int reg_1_idx = thread_idx / 8;
+  int reg_2_idx = thread_idx % 8;
+
+  float reg_1[8];
+  float reg_2[8];
+
+  float reg_array[16];
+
+  for (int dim_stride = 0; dim_stride < dim; dim_stride = dim_stride + 64) {
+
+    #pragma unroll
+    for (int i = 0; i < 16; i++) {
+      B_buffer[i * 128 + thread_idx] = dense_B_pt[dim_stride * 32 + i * 128 + thread_idx];
+    }
+
+    #pragma unroll
+    for (int i = 0; i < 16; i++) {
+      reg_array[i] = 0;
+    }
+
+    __syncthreads();
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      reg_1[i] = B_buffer[(reg_1_idx * 4 + i) * 32];
+      reg_2[i] = A_buffer[reg_2_idx * 4 + i];
+    }
+
+    #pragma unroll
+    for (int mini_dim_idx = 1; mini_dim_idx < 32; mini_dim_idx++) {
+      #pragma unroll
+      for (int i = 0; i < 4; i++) {
+        reg_1[(mini_dim_idx % 2) * 4 + i] = B_buffer[(reg_1_idx * 4 + i) * 32 + mini_dim_idx];
+        reg_2[(mini_dim_idx % 2) * 4 + i] = A_buffer[mini_dim_idx * 32 + reg_2_idx * 4 + i];
+      }
+      #pragma unroll
+      for (int i = 0; i < 4; i++) {
+        #pragma unroll
+        for (int j = 0; j < 4; j++) {
+          reg_array[i * 4 + j] += reg_1[((mini_dim_idx - 1) % 2) * 4 + i] * reg_2[((mini_dim_idx - 1) % 2) * 4 + j];
+        }
+      }
+    }
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      #pragma unroll
+      for (int j = 0; j < 4; j++) {
+        reg_array[i * 4 + j] += reg_1[4 + i] * reg_2[4 + j];
+      }
+    }
+
+    __syncthreads();
+
+    float *C_buffer = &buffer[threadIdx.y * 3072 + 1024]; // [64, 32]
+
+    #pragma unroll
+    for (int i = 0; i < 4; i++) {
+      #pragma unroll
+      for (int j = 0; j < 4; j++) {
+        C_buffer[(reg_1_idx * 4 + i) * 32 + reg_2_idx * 4 + j] = reg_array[i * 4 + j];
+      }
+    }
+    __syncthreads();
+
+    #pragma unroll
+    for (int i = 0; i < 16; i++) {
+      atomicAdd(&dense_C_pt[dim_stride * 32 + i * 128 + thread_idx], C_buffer[i * 128 + thread_idx]);
+    }
+    __syncthreads();
+
+  }
+
+}
+
+
+__global__ void reduce_sum_cuda_kernel(
+  float *sparse_A,  // [batch_size, num_block, 32, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *dense_C,   // [batch_size, A_num_block, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+) {
+
+  long batch_idx = blockIdx.y;
+  long block_idx = blockIdx.x * blockDim.y + threadIdx.y;
+
+  long thread_idx = threadIdx.x;
+
+  long batch_idx__block_idx = batch_idx * num_block + block_idx;
+
+  long AB_block_idx = indices[batch_idx__block_idx];
+  float *sparse_A_pt = &sparse_A[batch_idx__block_idx * 1024];
+
+  float reg_array[16];
+  float value = 0;
+
+  #pragma unroll
+  for (int i = 0; i < 8; i++) {
+    reg_array[i] = sparse_A_pt[i * 32 + thread_idx];
+  }
+  #pragma unroll
+  for (int stride = 8; stride < 32; stride = stride + 8) {
+    #pragma unroll
+    for (int i = 0; i < 8; i++) {
+      reg_array[(stride + i) % 16] = sparse_A_pt[(stride + i) * 32 + thread_idx];
+    }
+    #pragma unroll
+    for (int i = 0; i < 8; i++) {
+      value = value + reg_array[(stride - 8 + i) % 16];
+    }
+  }
+  #pragma unroll
+  for (int i = 0; i < 8; i++) {
+    value = value + reg_array[8 + i];
+  }
+
+  float *dense_C_pt = &dense_C[(batch_idx * A_num_block + AB_block_idx / B_num_block) * 32];
+
+  atomicAdd(&dense_C_pt[thread_idx], value);
+
+}
+
+__global__ void scatter_cuda_kernel(
+  float *dense_A,   // [batch_size, A_num_block, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *sparse_C,  // [batch_size, num_block, 32, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+) {
+
+  long batch_idx = blockIdx.y;
+  long block_idx = blockIdx.x * blockDim.y + threadIdx.y;
+
+  long thread_idx = threadIdx.x;
+
+  long batch_idx__block_idx = batch_idx * num_block + block_idx;
+
+  long AB_block_idx = indices[batch_idx__block_idx];
+  float *dense_A_pt = &dense_A[(batch_idx * A_num_block + AB_block_idx / B_num_block) * 32];
+  float *sparse_C_pt = &sparse_C[(batch_idx * num_block + block_idx) * 1024];
+
+  float value = dense_A_pt[thread_idx];
+
+  #pragma unroll
+  for (int i = 0; i < 32; i++) {
+    sparse_C_pt[i * 32 + thread_idx] = value;
+  }
+
+}

phivenv/Lib/site-packages/transformers/kernels/mra/cuda_kernel.h

@@ -0,0 +1,59 @@
+
+#define WARP_SIZE 32
+#define FULL_MASK 0xffffffff
+#define OPTIMAL_THREADS 256
+
+__global__ void index_max_cuda_kernel(
+  float *index_vals,       // [batch_size, 32, num_block]
+  int   *indices,        // [batch_size, num_block]
+  float *max_vals,        // [batch_size, A_num_block * 32]
+  float *max_vals_scatter,   // [batch_size, 32, num_block]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+);
+
+__global__ void mm_to_sparse_cuda_kernel(
+  float *dense_A,   // [batch_size, A_num_block, dim, 32]
+  float *dense_B,   // [batch_size, B_num_block, dim, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *sparse_C,  // [batch_size, num_block, 32, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long dim,
+  long num_block
+);
+
+__global__ void sparse_dense_mm_cuda_kernel(
+  float *sparse_A,  // [batch_size, num_block, 32, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *dense_B,   // [batch_size, B_num_block, dim, 32]
+  float *dense_C,   // [batch_size, A_num_block, dim, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long dim,
+  long num_block
+);
+
+__global__ void reduce_sum_cuda_kernel(
+  float *sparse_A,  // [batch_size, num_block, 32, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *dense_C,   // [batch_size, A_num_block, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+);
+
+__global__ void scatter_cuda_kernel(
+  float *dense_A,   // [batch_size, A_num_block, 32]
+  int   *indices,   // [batch_size, num_block]
+  float *sparse_C,  // [batch_size, num_block, 32, 32]
+  long batch_size,
+  long A_num_block,
+  long B_num_block,
+  long num_block
+);

phivenv/Lib/site-packages/transformers/kernels/mra/cuda_launch.cu

@@ -0,0 +1,154 @@
+#include <torch/extension.h>
+#include <ATen/ATen.h>
+#include "cuda_launch.h"
+#include "cuda_kernel.h"
+#include <vector>
+
+//////////////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////////
+
+std::vector<at::Tensor> index_max_kernel(
+  at::Tensor index_vals,  // [batch_size, 32, num_block]
+  at::Tensor indices,     // [batch_size, num_block],
+  int A_num_block,
+  int B_num_block
+) {
+  int batch_size = indices.size(0);
+  int num_block = indices.size(1);
+
+  at::Tensor max_vals = at::zeros({batch_size, A_num_block * 32}, index_vals.options());
+  at::Tensor max_vals_scatter = at::zeros({batch_size, 32, num_block}, index_vals.options());
+
+  dim3 threads(256);
+  dim3 blocks(batch_size);
+  int shared_mem = A_num_block * 32 * sizeof(float);
+
+  index_max_cuda_kernel<<<blocks, threads, shared_mem>>>(
+    index_vals.data_ptr<float>(),
+    indices.data_ptr<int>(),
+    max_vals.data_ptr<float>(),
+    max_vals_scatter.data_ptr<float>(),
+    batch_size,
+    A_num_block,
+    B_num_block,
+    num_block
+  );
+
+  return {max_vals, max_vals_scatter};
+}
+
+at::Tensor mm_to_sparse_kernel(
+  at::Tensor dense_A,  // [batch_size, A_num_block, dim, 32]
+  at::Tensor dense_B,  // [batch_size, B_num_block, dim, 32]
+  at::Tensor indices   // [batch_size, num_block]
+) {
+  int batch_size = dense_A.size(0);
+  int A_num_block = dense_A.size(1);
+  int B_num_block = dense_B.size(1);
+  int dim = dense_A.size(2);
+  int num_block = indices.size(1);
+
+  at::Tensor sparse_C = at::zeros({batch_size, num_block, 32, 32}, dense_A.options());
+
+  dim3 threads(64, 4);
+  dim3 blocks(num_block / 4, batch_size);
+
+  mm_to_sparse_cuda_kernel<<<blocks, threads>>>(
+    dense_A.data_ptr<float>(),
+    dense_B.data_ptr<float>(),
+    indices.data_ptr<int>(),
+    sparse_C.data_ptr<float>(),
+    batch_size,
+    A_num_block,
+    B_num_block,
+    dim,
+    num_block
+  );
+
+  return sparse_C;
+}
+
+at::Tensor sparse_dense_mm_kernel(
+  at::Tensor sparse_A,  // [batch_size, num_block, 32, 32]
+  at::Tensor indices,   // [batch_size, num_block]
+  at::Tensor dense_B,   // [batch_size, B_num_block, dim, 32]
+  int A_num_block
+) {
+  int batch_size = sparse_A.size(0);
+  int num_block = sparse_A.size(1);
+  int B_num_block = dense_B.size(1);
+  int dim = dense_B.size(2);
+
+  at::Tensor dense_C = at::zeros({batch_size, A_num_block, dim, 32}, dense_B.options());
+
+  dim3 threads(128, 2);
+  dim3 blocks(num_block / 2, batch_size);
+
+  sparse_dense_mm_cuda_kernel<<<blocks, threads>>>(
+    sparse_A.data_ptr<float>(),
+    indices.data_ptr<int>(),
+    dense_B.data_ptr<float>(),
+    dense_C.data_ptr<float>(),
+    batch_size,
+    A_num_block,
+    B_num_block,
+    dim,
+    num_block
+  );
+
+  return dense_C;
+}
+
+at::Tensor reduce_sum_kernel(
+  at::Tensor sparse_A,  // [batch_size, num_block, 32, 32]
+  at::Tensor indices,   // [batch_size, num_block]
+  int A_num_block,
+  int B_num_block
+) {
+  int batch_size = sparse_A.size(0);
+  int num_block = sparse_A.size(1);
+
+  at::Tensor dense_C = at::zeros({batch_size, A_num_block, 32}, sparse_A.options());
+
+  dim3 threads(32, 4);
+  dim3 blocks(num_block / 4, batch_size);
+
+  reduce_sum_cuda_kernel<<<blocks, threads>>>(
+    sparse_A.data_ptr<float>(),
+    indices.data_ptr<int>(),
+    dense_C.data_ptr<float>(),
+    batch_size,
+    A_num_block,
+    B_num_block,
+    num_block
+  );
+
+  return dense_C;
+}
+
+at::Tensor scatter_kernel(
+  at::Tensor dense_A,   // [batch_size, A_num_block, 32]
+  at::Tensor indices,   // [batch_size, num_block]
+  int B_num_block
+) {
+  int batch_size = dense_A.size(0);
+  int A_num_block = dense_A.size(1);
+  int num_block = indices.size(1);
+
+  at::Tensor sparse_C = at::zeros({batch_size, num_block, 32, 32}, dense_A.options());
+
+  dim3 threads(32, 4);
+  dim3 blocks(num_block / 4, batch_size);
+
+  scatter_cuda_kernel<<<blocks, threads>>>(
+    dense_A.data_ptr<float>(),
+    indices.data_ptr<int>(),
+    sparse_C.data_ptr<float>(),
+    batch_size,
+    A_num_block,
+    B_num_block,
+    num_block
+  );
+
+  return sparse_C;
+}

phivenv/Lib/site-packages/transformers/kernels/mra/cuda_launch.h

@@ -0,0 +1,39 @@
+#include <torch/extension.h>
+#include <ATen/ATen.h>
+#include <vector>
+
+#define min(a, b) ((a)<(b)?(a):(b))
+#define max(a, b) ((a)>(b)?(a):(b))
+
+std::vector<at::Tensor> index_max_kernel(
+  at::Tensor index_vals,
+  at::Tensor indices,
+  int A_num_block,
+  int B_num_block
+);
+
+at::Tensor mm_to_sparse_kernel(
+  at::Tensor dense_A,
+  at::Tensor dense_B,
+  at::Tensor indices
+);
+
+at::Tensor sparse_dense_mm_kernel(
+  at::Tensor sparse_A,
+  at::Tensor indices,
+  at::Tensor dense_B,
+  int A_num_block
+);
+
+at::Tensor reduce_sum_kernel(
+  at::Tensor sparse_A,
+  at::Tensor indices,
+  int A_num_block,
+  int B_num_block
+);
+
+at::Tensor scatter_kernel(
+  at::Tensor dense_A,
+  at::Tensor indices,
+  int B_num_block
+);

phivenv/Lib/site-packages/transformers/kernels/mra/torch_extension.cpp

@@ -0,0 +1,78 @@
+#include <torch/extension.h>
+#include <ATen/ATen.h>
+#include "cuda_launch.h"
+#include <vector>
+
+std::vector<at::Tensor> index_max(
+  at::Tensor index_vals,
+  at::Tensor indices,
+  int A_num_block,
+  int B_num_block
+) {
+  return index_max_kernel(
+    index_vals,
+    indices,
+    A_num_block,
+    B_num_block
+  );
+}
+
+at::Tensor mm_to_sparse(
+  at::Tensor dense_A,
+  at::Tensor dense_B,
+  at::Tensor indices
+) {
+  return mm_to_sparse_kernel(
+    dense_A,
+    dense_B,
+    indices
+  );
+}
+
+at::Tensor sparse_dense_mm(
+  at::Tensor sparse_A,
+  at::Tensor indices,
+  at::Tensor dense_B,
+  int A_num_block
+) {
+  return sparse_dense_mm_kernel(
+    sparse_A,
+    indices,
+    dense_B,
+    A_num_block
+  );
+}
+
+at::Tensor reduce_sum(
+  at::Tensor sparse_A,
+  at::Tensor indices,
+  int A_num_block,
+  int B_num_block
+) {
+  return reduce_sum_kernel(
+    sparse_A,
+    indices,
+    A_num_block,
+    B_num_block
+  );
+}
+
+at::Tensor scatter(
+  at::Tensor dense_A,
+  at::Tensor indices,
+  int B_num_block
+) {
+  return scatter_kernel(
+    dense_A,
+    indices,
+    B_num_block
+  );
+}
+
+PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
+  m.def("index_max", &index_max, "index_max (CUDA)");
+  m.def("mm_to_sparse", &mm_to_sparse, "mm_to_sparse (CUDA)");
+  m.def("sparse_dense_mm", &sparse_dense_mm, "sparse_dense_mm (CUDA)");
+  m.def("reduce_sum", &reduce_sum, "reduce_sum (CUDA)");
+  m.def("scatter", &scatter, "scatter (CUDA)");
+}

phivenv/Lib/site-packages/transformers/kernels/rwkv/wkv_cuda.cu

@@ -0,0 +1,187 @@
+#include <stdio.h>
+#include <assert.h>
+
+#define MIN_VALUE (-1e38)
+
+template <typename F>
+__global__ void kernel_forward(
+    const int B, const int T, const int C, const F *__restrict__ const _w, const F *__restrict__ const _u,
+    const F *__restrict__ const _k, const F *__restrict__ const _v, F *__restrict__ const _y
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset = _b * T * C + _c;
+
+    F u = _u[_c];
+    F w = _w[_c];
+    const F *__restrict__ const k = _k + _offset;
+    const F *__restrict__ const v = _v + _offset;
+    F *__restrict__ const y = _y + _offset;
+
+    // aa and bb are running sums divided by exp(pp) (to avoid overflow)
+    F aa = 0, bb = 0, pp = MIN_VALUE;
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const F kk = k[ii];
+        const F vv = v[ii];
+
+        F ww = u + kk;
+        F p = max(pp, ww);
+        F e1 = exp(pp - p);
+        F e2 = exp(ww - p);
+        y[ii] = (e1 * aa + e2 * vv) / (e1 * bb + e2);
+        
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+}
+
+template <typename F>
+__global__ void kernel_forward_with_state(
+    const int B, const int T, const int C, const F *__restrict__ const _w, const F *__restrict__ const _u,
+    const F *__restrict__ const _k, const F *__restrict__ const _v, F *__restrict__ const _y, F *__restrict__ const _s
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset_s = _b * C * 3 + _c * 3;
+    const int _offset = _b * T * C + _c;
+
+    F u = _u[_c];
+    F w = _w[_c];
+    const F *__restrict__ const k = _k + _offset;
+    const F *__restrict__ const v = _v + _offset;
+    F *__restrict__ const y = _y + _offset;
+    F *__restrict__ const s = _s + _offset_s;
+
+    // aa and bb are running sums divided by exp(pp) (to avoid overflow)
+    F aa = s[0], bb = s[1], pp = s[2];
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const F kk = k[ii];
+        const F vv = v[ii];
+
+        F ww = u + kk;
+        F p = max(pp, ww);
+        F e1 = exp(pp - p);
+        F e2 = exp(ww - p);
+        y[ii] = (e1 * aa + e2 * vv) / (e1 * bb + e2);
+        
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+    s[0] = aa;
+    s[1] = bb;
+    s[2] = pp;
+}
+
+template <typename F>
+__global__ void kernel_backward(
+    const int B, const int T, const int C, const F *__restrict__ const _w, const F *__restrict__ const _u,
+    const F *__restrict__ const _k, const F *__restrict__ const _v, const F *__restrict__ const _y,
+    const F *__restrict__ const _gy, F *__restrict__ const _gw, F *__restrict__ const _gu, F *__restrict__ const _gk,
+    F *__restrict__ const _gv
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset = _b * T * C + _c;
+
+    F u = _u[_c];
+    F w = _w[_c];
+    const F *__restrict__ const k = _k + _offset;
+    const F *__restrict__ const v = _v + _offset;
+    const F *__restrict__ const y = _y + _offset;
+    const F *__restrict__ const gy = _gy + _offset;
+    F *__restrict__ const gk = _gk + _offset;
+    F *__restrict__ const gv = _gv + _offset;
+
+    F q[Tmax], r[Tmax];
+
+    F gw = 0, gu = 0, aa = 0, bb = 0, ga = 0, gb = 0, pp = MIN_VALUE;
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const F kk = k[ii];
+        const F vv = v[ii];
+        const F yy = y[ii];
+
+        F ww = u + kk;
+        F p = max(pp, ww);
+        F e1 = exp(pp - p);
+        F e2 = exp(ww - p);
+        const F qq = gy[ii] / (e1 * bb + e2);
+        gw += (ga - gb * yy) * e1 * qq;
+        gu += (vv - yy) * e2 * qq;
+        q[i] = qq;
+        r[i] = ww - p;
+
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        ga = e1 * (aa + ga);
+        gb = e1 * (bb + gb);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+    const int _offsetBC = _b * C + _c;
+    _gw[_offsetBC] = gw * _w[_c]; // multiply by w because of w -> -exp(w) in python forward()
+    _gu[_offsetBC] = gu;
+
+    aa = 0, bb = 0, pp = MIN_VALUE;
+    for (int i = T - 1; i >= 0; i--) {
+        const int ii = i * C;
+        const F kk = k[ii];
+        const F vv = v[ii];
+        const F yy = y[ii];
+        const F qq = q[i];
+        const F rr = r[i];
+
+        F e1 = qq * exp(rr);
+        F e2 = exp(kk + pp);
+        gk[ii] = e1 * (vv - yy) + e2 * (aa * vv + bb);
+        gv[ii] = e1 + e2 * aa;
+
+        const F ww = w + pp;
+        const F www = rr - u - kk;
+        const F p = max(ww, www);
+        e1 = exp(ww - p);
+        e2 = qq * exp(www - p);
+        aa = e1 * aa + e2;
+        bb = e1 * bb - e2 * yy;
+        pp = p;
+    }
+}
+
+void cuda_forward(int B, int T, int C, float *w, float *u, float *k, float *v, float *y) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_forward<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y);
+}
+
+void cuda_forward_with_state(int B, int T, int C, float *w, float *u, float *k, float *v, float *y, float *s) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_forward_with_state<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y, s);
+}
+
+void cuda_backward(int B, int T, int C, float *w, float *u, float *k, float *v, float *y, float *gy, float *gw, float *gu, float *gk, float *gv) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_backward<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y, gy, gw, gu, gk, gv);
+}

phivenv/Lib/site-packages/transformers/kernels/rwkv/wkv_cuda_bf16.cu

@@ -0,0 +1,186 @@
+#include <stdio.h>
+#include <assert.h>
+#include "ATen/ATen.h"
+#define MIN_VALUE (-1e38)
+typedef at::BFloat16 bf16;
+
+__global__ void kernel_forward_bf16(
+    const int B, const int T, const int C, const float *__restrict__ const _w, const bf16 *__restrict__ const _u,
+    const bf16 *__restrict__ const _k, const bf16 *__restrict__ const _v, bf16 *__restrict__ const _y
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset = _b * T * C + _c;
+
+    float u = float(_u[_c]);
+    float w = _w[_c];
+    const bf16 *__restrict__ const k = _k + _offset;
+    const bf16 *__restrict__ const v = _v + _offset;
+    bf16 *__restrict__ const y = _y + _offset;
+
+    // aa and bb are running sums divided by exp(pp) (to avoid overflow)
+    float aa = 0, bb = 0, pp = MIN_VALUE;
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const float kk = float(k[ii]);
+        const float vv = float(v[ii]);
+
+        float ww = u + kk;
+        float p = max(pp, ww);
+        float e1 = exp(pp - p);
+        float e2 = exp(ww - p);
+        y[ii] = bf16((e1 * aa + e2 * vv) / (e1 * bb + e2));
+        
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+}
+
+__global__ void kernel_forward_with_state_bf16(
+    const int B, const int T, const int C, const float *__restrict__ const _w, const bf16 *__restrict__ const _u,
+    const bf16 *__restrict__ const _k, const bf16 *__restrict__ const _v, bf16 *__restrict__ const _y,
+    float *__restrict__ const _s
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset_s = _b * C * 3 + _c * 3;
+    const int _offset = _b * T * C + _c;
+
+    float u = float(_u[_c]);
+    float w = _w[_c];
+    const bf16 *__restrict__ const k = _k + _offset;
+    const bf16 *__restrict__ const v = _v + _offset;
+    bf16 *__restrict__ const y = _y + _offset;
+    float *__restrict__ const s = _s + _offset_s;
+
+    // aa and bb are running sums divided by exp(pp) (to avoid overflow)
+    float aa = s[0], bb = s[1], pp = s[2];
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const float kk = float(k[ii]);
+        const float vv = float(v[ii]);
+
+        float ww = u + kk;
+        float p = max(pp, ww);
+        float e1 = exp(pp - p);
+        float e2 = exp(ww - p);
+        y[ii] = bf16(e1 * aa + e2 * vv) / (e1 * bb + e2);
+        
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+    s[0] = aa;
+    s[1] = bb;
+    s[2] = pp;
+}
+
+__global__ void kernel_backward_bf16(
+    const int B, const int T, const int C, const float *__restrict__ const _w, const bf16 *__restrict__ const _u,
+    const bf16 *__restrict__ const _k, const bf16 *__restrict__ const _v, const bf16 *__restrict__ const _y,
+    const bf16 *__restrict__ const _gy, bf16 *__restrict__ const _gw, bf16 *__restrict__ const _gu,
+    bf16 *__restrict__ const _gk, bf16 *__restrict__ const _gv
+) {
+    const int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    const int _b = idx / C;
+    const int _c = idx % C;
+    const int _offset = _b * T * C + _c;
+
+    float u = float(_u[_c]);
+    float w = _w[_c];
+    const bf16 *__restrict__ const k = _k + _offset;
+    const bf16 *__restrict__ const v = _v + _offset;
+    const bf16 *__restrict__ const y = _y + _offset;
+    const bf16 *__restrict__ const gy = _gy + _offset;
+    bf16 *__restrict__ const gk = _gk + _offset;
+    bf16 *__restrict__ const gv = _gv + _offset;
+
+    float q[Tmax], r[Tmax];
+
+    float gw = 0, gu = 0, aa = 0, bb = 0, ga = 0, gb = 0, pp = MIN_VALUE;
+    for (int i = 0; i < T; i++) {
+        const int ii = i * C;
+        const float kk = float(k[ii]);
+        const float vv = float(v[ii]);
+        const float yy = float(y[ii]);
+
+        float ww = u + kk;
+        float p = max(pp, ww);
+        float e1 = exp(pp - p);
+        float e2 = exp(ww - p);
+        const float qq = float(gy[ii]) / (e1 * bb + e2);
+        gw += (ga - gb * yy) * e1 * qq;
+        gu += (vv - yy) * e2 * qq;
+        q[i] = qq;
+        r[i] = ww - p;
+
+        ww = w + pp;
+        p = max(ww, kk);
+        e1 = exp(ww - p);
+        e2 = exp(kk - p);
+        ga = e1 * (aa + ga);
+        gb = e1 * (bb + gb);
+        aa = e1 * aa + e2 * vv;
+        bb = e1 * bb + e2;
+        pp = p;
+    }
+    const int _offsetBC = _b * C + _c;
+    _gw[_offsetBC] = bf16(gw * _w[_c]); // multiply by w because of w -> -exp(w) in python forward()
+    _gu[_offsetBC] = bf16(gu);
+
+    aa = 0, bb = 0, pp = MIN_VALUE;
+    for (int i = T - 1; i >= 0; i--) {
+        const int ii = i * C;
+        const float kk = float(k[ii]);
+        const float vv = float(v[ii]);
+        const float yy = float(y[ii]);
+        const float qq = q[i];
+        const float rr = r[i];
+
+        float e1 = qq * exp(rr);
+        float e2 = exp(kk + pp);
+        gk[ii] = bf16(e1 * (vv - yy) + e2 * (aa * vv + bb));
+        gv[ii] = bf16(e1 + e2 * aa);
+
+        const float ww = w + pp;
+        const float www = rr - u - kk;
+        const float p = max(ww, www);
+        e1 = exp(ww - p);
+        e2 = qq * exp(www - p);
+        aa = e1 * aa + e2;
+        bb = e1 * bb - e2 * yy;
+        pp = p;
+    }
+}
+
+void cuda_forward_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_forward_bf16<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y);
+}
+
+void cuda_forward_with_state_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y, float *s) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_forward_with_state_bf16<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y, s);
+}
+
+void cuda_backward_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y, bf16 *gy, bf16 *gw, bf16 *gu, bf16 *gk, bf16 *gv) {
+    dim3 threadsPerBlock( min(C, 32) ); // requires --maxrregcount 60 for optimal performance
+    assert(B * C % threadsPerBlock.x == 0);
+    dim3 numBlocks(B * C / threadsPerBlock.x);
+    kernel_backward_bf16<<<numBlocks, threadsPerBlock>>>(B, T, C, w, u, k, v, y, gy, gw, gu, gk, gv);
+}

phivenv/Lib/site-packages/transformers/kernels/rwkv/wkv_op.cpp

@@ -0,0 +1,66 @@
+#include <torch/extension.h>
+#include "ATen/ATen.h"
+typedef at::BFloat16 bf16;
+
+void cuda_forward(int B, int T, int C, float *w, float *u, float *k, float *v, float *y);
+void cuda_forward_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y);
+void cuda_forward_with_state(int B, int T, int C, float *w, float *u, float *k, float *v, float *y, float *s);
+void cuda_forward_with_state_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y, float *s);
+void cuda_backward(int B, int T, int C, float *w, float *u, float *k, float *v, float *y, float *gy, float *gw, float *gu, float *gk, float *gv);
+void cuda_backward_bf16(int B, int T, int C, float *w, bf16 *u, bf16 *k, bf16 *v, bf16 *y, bf16 *gy, bf16 *gw, bf16 *gu, bf16 *gk, bf16 *gv);
+
+void forward(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_forward(B, T, C, w.data_ptr<float>(), u.data_ptr<float>(), k.data_ptr<float>(), v.data_ptr<float>(), y.data_ptr<float>());
+}
+void forward_bf16(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_forward_bf16(B, T, C, w.data_ptr<float>(), u.data_ptr<bf16>(), k.data_ptr<bf16>(), v.data_ptr<bf16>(), y.data_ptr<bf16>());
+}
+void forward_with_state(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y, torch::Tensor &s) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_forward_with_state(B, T, C, w.data_ptr<float>(), u.data_ptr<float>(), k.data_ptr<float>(), v.data_ptr<float>(), y.data_ptr<float>(), s.data_ptr<float>());
+}
+void forward_with_state_bf16(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y, torch::Tensor &s) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_forward_with_state_bf16(B, T, C, w.data_ptr<float>(), u.data_ptr<bf16>(), k.data_ptr<bf16>(), v.data_ptr<bf16>(), y.data_ptr<bf16>(), s.data_ptr<float>());
+}
+void backward(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y, torch::Tensor &gy, torch::Tensor &gw, torch::Tensor &gu, torch::Tensor &gk, torch::Tensor &gv) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_backward(B, T, C, w.data_ptr<float>(), u.data_ptr<float>(), k.data_ptr<float>(), v.data_ptr<float>(), y.data_ptr<float>(), gy.data_ptr<float>(), gw.data_ptr<float>(), gu.data_ptr<float>(), gk.data_ptr<float>(), gv.data_ptr<float>());
+}
+void backward_bf16(torch::Tensor &w, torch::Tensor &u, torch::Tensor &k, torch::Tensor &v, torch::Tensor &y, torch::Tensor &gy, torch::Tensor &gw, torch::Tensor &gu, torch::Tensor &gk, torch::Tensor &gv) {
+    const int B = k.size(0);
+    const int T = k.size(1);
+    const int C = k.size(2);
+    cuda_backward_bf16(B, T, C, w.data_ptr<float>(), u.data_ptr<bf16>(), k.data_ptr<bf16>(), v.data_ptr<bf16>(), y.data_ptr<bf16>(),
+        gy.data_ptr<bf16>(), gw.data_ptr<bf16>(), gu.data_ptr<bf16>(), gk.data_ptr<bf16>(), gv.data_ptr<bf16>());
+}
+
+PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
+    m.def("forward", &forward, "wkv forward");
+    m.def("forward_bf16", &forward_bf16, "wkv forward bf16");
+    m.def("forward_with_state", &forward_with_state, "wkv forward with state");
+    m.def("forward_with_state_bf16", &forward_with_state_bf16, "wkv forward with state bf16");
+    m.def("backward", &backward, "wkv backward");
+    m.def("backward_bf16", &backward_bf16, "wkv backward bf16");
+}
+
+TORCH_LIBRARY(wkv, m) {
+    m.def("forward", forward);
+    m.def("forward_bf16", forward_bf16);
+    m.def("forward_with_state", forward_with_state);
+    m.def("forward_with_state_bf16", forward_with_state_bf16);
+    m.def("backward", backward);
+    m.def("backward_bf16", backward_bf16);
+}

phivenv/Lib/site-packages/transformers/kernels/yoso/common.h

@@ -0,0 +1,10 @@
+
+#define min(a, b) ((a)<(b)?(a):(b))
+#define max(a, b) ((a)>(b)?(a):(b))
+#define ceil_divide(a, b) ((a)/(b)+((a)%(b)!=0))
+#define select(cond, a, b) ((cond)?(a):(b))
+#define PI 3.141592
+#define EPSILON 1e-8
+#define MAX_VAL 1e12
+#define MIN_VAL -1e12
+#define EMPTY_VALUE -1