!3800 add gpu multinomial backend

Merge pull request !3800 from baihuawei/multinomial-c

mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/multinomial_impl.cu

+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#include "multinomial_impl.cuh"
+
+template <typename T>
+__global__ void NormInput(T *input, const size_t distributions, const size_t categories) {
+  size_t size = distributions * categories;
+  for (size_t pos = blockIdx.x * blockDim.x + threadIdx.x; pos < (size); pos += blockDim.x * gridDim.x) {
+    if ((pos + 1) % categories != 0) {
+      int de_pos = (1 + pos / categories) * categories - 1;
+      input[pos] /= input[de_pos];
+    }
+  }
+  return;
+}
+
+template <typename T>
+__global__ void CheckZeroKernel(const size_t distributions, const size_t categories, const T *input, T *out) {
+  out[0] = 0;
+  for (size_t pos = blockIdx.x * blockDim.x + threadIdx.x; pos < (distributions); pos += blockDim.x * gridDim.x) {
+    if (input[(1 + pos) * categories - 1] <= 0) {
+      out[0] = 1;
+    }
+  }
+  return;
+}
+
+template <typename T>
+void CheckZero(const size_t distributions, const size_t categories, const T *input, T *output,
+               cudaStream_t cuda_stream) {
+  CheckZeroKernel<<<GET_BLOCKS(distributions), GET_THREADS, 0, cuda_stream>>>(distributions, categories, input, output);
+}
+
+template <typename T>
+__global__ void CheckNonNegKernel(const size_t size, const T *input, T *out) {
+  out[0] = 0;
+  for (size_t pos = blockIdx.x * blockDim.x + threadIdx.x; pos < (size); pos += blockDim.x * gridDim.x) {
+    if (input[pos] < 0) {
+      out[0] = 1;
+    }
+  }
+  return;
+}
+
+template <typename T>
+void CheckNonNeg(const size_t size, const T *input, T *output, cudaStream_t cuda_stream) {
+  CheckNonNegKernel<<<GET_BLOCKS(size), GET_THREADS, 0, cuda_stream>>>(size, input, output);
+}
+
+template <typename T>
+__device__ int BinarySearchForMultinomial(T *start_addr, int size, T rand) {
+  int start = 0;
+  int end = size;
+  while (end - start > 0) {
+    int mid = start + (end - start) / 2;
+    T mid_val = start_addr[mid];
+    if (mid_val < rand) {
+      start = mid + 1;
+    } else {
+      end = mid;
+    }
+  }
+  if (start == size) {
+    start = size - 1;
+  }
+  return start;
+}
+
+template <typename T>
+__global__ void MultinomialKernel(int seed, T *input, int num_sample, curandState *globalState, int *output,
+                                  size_t distributions, size_t categories) {
+  int count = num_sample * distributions;
+  for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < count; i += blockDim.x * gridDim.x) {
+    int j = i / num_sample % distributions;
+    curand_init(seed, i, 0, &globalState[i]);
+    auto rand = curand_uniform(&globalState[i]);
+    int pick = BinarySearchForMultinomial(input + j * categories, categories, rand);
+    output[i] = pick;
+  }
+  return;
+}
+
+template <typename T>
+void Multinomial(int seed, T *input, int num_sample, curandState *globalState, int *output, size_t distributions,
+                 size_t categories, cudaStream_t cuda_stream) {
+  int RNG_seed = 0;
+  if (seed != 0) {
+    RNG_seed = seed;
+  } else {
+    RNG_seed = time(NULL);
+  }
+  int count = distributions * num_sample;
+  int count1 = distributions * categories;
+  NormInput<<<GET_BLOCKS(count1), GET_THREADS, 0, cuda_stream>>>(input, distributions, categories);
+  MultinomialKernel<<<GET_BLOCKS(count), GET_THREADS, 0, cuda_stream>>>(RNG_seed, input, num_sample, globalState,
+                                                                        output, distributions, categories);
+  return;
+}
+
+template void Multinomial<float>(int seed, float *input, int num_sample, curandState *globalState, int *output,
+                                 size_t distributions, size_t categories, cudaStream_t cuda_stream);
+template void CheckNonNeg<float>(const size_t size, const float *input, float *output, cudaStream_t cuda_stream);
+template void CheckZero<float>(const size_t distributions, const size_t categories, const float *input, float *output,
+                               cudaStream_t cuda_stream);

mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/multinomial_impl.cuh

+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_IMPL_CUH_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_IMPL_CUH_
+#include <curand_kernel.h>
+#include "runtime/device/gpu/cuda_common.h"
+
+template <typename T>
+void Multinomial(int seed, T *input, int num_sample, curandState *globalState, int *output, size_t distributions,
+                 size_t categories, cudaStream_t cuda_stream);
+template <typename T>
+void CheckNonNeg(const size_t size, const T *input, T *output, cudaStream_t stream);
+template <typename T>
+void CheckZero(const size_t distributions, const size_t categories, const T *input, T *output, cudaStream_t stream);
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_IMPL_CUH_

mindspore/ccsrc/backend/kernel_compiler/gpu/math/multinomial_gpu_kernel.cc

+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#include "backend/kernel_compiler/gpu/math/multinomial_gpu_kernel.h"
+
+namespace mindspore {
+namespace kernel {
+MS_REG_GPU_KERNEL_ONE(
+  Multinomial,
+  KernelAttr().AddInputAttr(kNumberTypeFloat32).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeInt32),
+  MultinomialGpuKernel, float)
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/gpu/math/multinomial_gpu_kernel.h

+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_GPU_KERNEL_H_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_GPU_KERNEL_H_
+
+#include <curand_kernel.h>
+#include <cuda_runtime_api.h>
+#include <vector>
+#include <string>
+#include <map>
+#include "backend/kernel_compiler/gpu/gpu_kernel.h"
+#include "backend/kernel_compiler/gpu/gpu_kernel_factory.h"
+#include "backend/kernel_compiler/gpu/cuda_impl/multinomial_impl.cuh"
+#include "backend/kernel_compiler/gpu/cuda_impl/float_status_impl.cuh"
+#include "backend/kernel_compiler/gpu/cuda_impl/cumsum_impl.cuh"
+
+namespace mindspore {
+namespace kernel {
+template <typename T>
+class MultinomialGpuKernel : public GpuKernel {
+ public:
+  MultinomialGpuKernel() : input_size_0_(0), output_size_(0), distributions_(0), workspace_size_(sizeof(curandState)) {}
+  ~MultinomialGpuKernel() override = default;
+
+  const std::vector<size_t> &GetInputSizeList() const override { return input_size_list_; }
+  const std::vector<size_t> &GetOutputSizeList() const override { return output_size_list_; }
+  const std::vector<size_t> &GetWorkspaceSizeList() const override { return workspace_size_list_; }
+
+  bool Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+              const std::vector<AddressPtr> &outputs, void *stream_ptr) override {
+    void *workspace_addr = GetDeviceAddress<void *>(workspace, 0);
+    curandState *devStates = reinterpret_cast<curandState *>(workspace_addr);
+    int *output_addr = GetDeviceAddress<int>(outputs, 0);
+    T *input_addr = GetDeviceAddress<T>(inputs, 0);
+    int categories = SizeToInt(inputs[0]->size / sizeof(T)) / distributions_;
+    int num_sample = SizeToInt(outputs[0]->size / sizeof(T)) / distributions_;
+    // check input
+    T *flag = nullptr;
+    T *cflag = nullptr;
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMalloc(reinterpret_cast<void **>(&cflag), sizeof(T)), "cudaMalloc failed.");
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMallocHost(&flag, sizeof(T)), "cudaMallocHost failed.");
+    CalFloatStatus(input_size_0_ / sizeof(T), input_addr, cflag, reinterpret_cast<cudaStream_t>(stream_ptr));
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMemcpy(flag, cflag, sizeof(T), cudaMemcpyDeviceToHost), "cudaMemcpyAsync failed.");
+    if (*flag > 0) {
+      MS_LOG(EXCEPTION) << "Input is invalid (containing NaN, -inf or inf)";
+    }
+    CheckNonNeg(input_size_0_ / sizeof(T), input_addr, cflag, reinterpret_cast<cudaStream_t>(stream_ptr));
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMemcpy(flag, cflag, sizeof(T), cudaMemcpyDeviceToHost), "cudaMemcpyAsync failed.");
+    if (*flag > 0) {
+      MS_LOG(EXCEPTION) << "Input is invalid (input element < 0)";
+    }
+    T *cum_sum_input = nullptr;
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMalloc(reinterpret_cast<void **>(&cum_sum_input), input_size_0_),
+                               "cudaMalloc failed.");
+    CumSum(input_addr, cum_sum_input, cum_sum_input, IntToSize(distributions_), IntToSize(categories), 1,
+           IntToSize(categories), 1, false, false, reinterpret_cast<cudaStream_t>(stream_ptr));
+    CheckZero(IntToSize(distributions_), IntToSize(categories), cum_sum_input, cflag,
+              reinterpret_cast<cudaStream_t>(stream_ptr));
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMemcpy(flag, cflag, sizeof(T), cudaMemcpyDeviceToHost), "cudaMemcpyAsync failed.");
+    if (*flag > 0) {
+      MS_LOG(EXCEPTION) << "Input is invalid (sum <= 0)";
+    }
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaStreamSynchronize(reinterpret_cast<cudaStream_t>(stream_ptr)),
+                               "cudaStreamSynchronize failed.");
+    Multinomial(seed_, cum_sum_input, num_sample, devStates, output_addr, IntToSize(distributions_),
+                IntToSize(categories), reinterpret_cast<cudaStream_t>(stream_ptr));
+
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaFree(cum_sum_input), "cudaFree failed.");
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaFree(cflag), "cudaFree failed.");
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaFreeHost(flag), "cudaFreeHost failed.");
+    return true;
+  }
+  bool Init(const CNodePtr &kernel_node) override {
+    std::string kernel_name = AnfAlgo::GetCNodeName(kernel_node);
+    size_t input_num = AnfAlgo::GetInputTensorNum(kernel_node);
+    if (input_num != 2) {
+      MS_LOG(ERROR) << "Input number is " << input_num << ", but multinomial needs 2 input.";
+      return false;
+    }
+    size_t output_num = AnfAlgo::GetOutputTensorNum(kernel_node);
+    if (output_num != 1) {
+      MS_LOG(ERROR) << "Output number is " << output_num << ", but multinomial needs 1 output.";
+      return false;
+    }
+    auto input_shape_0 = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
+    if (input_shape_0.size() == 1) {
+      distributions_ = 1;
+    } else {
+      distributions_ = input_shape_0[0];
+    }
+    input_size_0_ = sizeof(T);
+    for (size_t i = 0; i < input_shape_0.size(); i++) {
+      input_size_0_ *= input_shape_0[i];
+    }
+    auto output_shape = AnfAlgo::GetOutputInferShape(kernel_node, 0);
+    output_size_ = sizeof(int);
+    for (size_t i = 0; i < output_shape.size(); i++) {
+      output_size_ *= output_shape[i];
+      workspace_size_ *= output_shape[i];
+    }
+    seed_ = GetValue<int>(AnfAlgo::GetCNodePrimitive(kernel_node)->GetAttr("seed"));
+    InitSizeLists();
+    return true;
+  }
+
+ protected:
+  void InitSizeLists() override {
+    input_size_list_.push_back(input_size_0_);
+    input_size_list_.push_back(sizeof(int));
+    output_size_list_.push_back(output_size_);
+    workspace_size_list_.push_back(workspace_size_);
+  }
+
+ private:
+  size_t input_size_0_;
+  size_t output_size_;
+  size_t distributions_;
+  size_t workspace_size_;
+  int seed_;
+  std::vector<size_t> input_size_list_;
+  std::vector<size_t> output_size_list_;
+  std::vector<size_t> workspace_size_list_;
+};
+}  // namespace kernel
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_MULTINOMIAL_GPU_KERNEL_H_

tests/st/ops/gpu/test_multinomial_op.py

+# Copyright 2020 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+import numpy as np
+import pytest
+from mindspore.ops import composite as C
+import mindspore.context as context
+from mindspore import Tensor
+
+context.set_context(device_target='GPU')
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_multinomial():
+    x0 = Tensor(np.array([0.9, 0.2]).astype(np.float32))
+    x1 = Tensor(np.array([[0.9, 0.2], [0.9, 0.2]]).astype(np.float32))
+    out0 = C.multinomial(x0, 1, True)
+    out1 = C.multinomial(x0, 2, True)
+    out2 = C.multinomial(x1, 6, True)
+    out3 = C.multinomial(x0, 1, False)
+    out4 = C.multinomial(x0, 2, False)
+    assert out0.asnumpy().shape == (1,)
+    assert out1.asnumpy().shape == (2,)
+    assert out2.asnumpy().shape == (2, 6)
+    assert out3.asnumpy().shape == (1,)
+    assert out4.asnumpy().shape == (2,)