From 69ec13cdf03b50e6b7602561af9977e06d235f1f Mon Sep 17 00:00:00 2001
From: pangyoki <pangyoki@126.com>
Date: Mon, 19 Oct 2020 19:21:42 +0800
Subject: [PATCH] cherry pick 27846 Add uniform_random XPU kernel, test=kunlun
 (#28057)

* support uniform_random op on Baidu Kunlun

* change dtype of attr shape from int to int64_t

* kunlun ci, test=kunlun

* new version, test=kunlun

* change boost_get to BOOST_GET_CONST

* change boost_get to BOOST_GET_CONST, test=kunlun

* use Generator to generate random number and optimize format

* run Kunlun CI, test=kunlun

* add TODO, test=kunlun
---
 .../fluid/operators/uniform_random_op_xpu.cc  | 75 +++++++++++++++++++
 .../xpu/test_uniform_random_op_xpu.py         | 51 +++++++++++++
 2 files changed, 126 insertions(+)
 create mode 100644 paddle/fluid/operators/uniform_random_op_xpu.cc
 create mode 100644 python/paddle/fluid/tests/unittests/xpu/test_uniform_random_op_xpu.py

diff --git a/paddle/fluid/operators/uniform_random_op_xpu.cc b/paddle/fluid/operators/uniform_random_op_xpu.cc
new file mode 100644
index 00000000000..507bd7e9ea9
--- /dev/null
+++ b/paddle/fluid/operators/uniform_random_op_xpu.cc
@@ -0,0 +1,75 @@
+/* Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+
+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. */
+
+#ifdef PADDLE_WITH_XPU
+
+#include "paddle/fluid/operators/uniform_random_op.h"
+#include <string>
+#include "paddle/fluid/framework/generator.h"
+#include "paddle/fluid/framework/op_registry.h"
+#include "paddle/fluid/framework/operator.h"
+
+namespace paddle {
+namespace operators {
+
+template <typename T>
+class XPUUniformRandomKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext &ctx) const override {
+    framework::Tensor *tensor = nullptr;
+    auto out_var = ctx.OutputVar("Out");
+    if (out_var->IsType<framework::LoDTensor>()) {
+      tensor = out_var->GetMutable<framework::LoDTensor>();
+    } else if (out_var->IsType<framework::SelectedRows>()) {
+      auto shape = ctx.Attr<std::vector<int64_t>>("shape");
+      auto *selected_rows = out_var->GetMutable<framework::SelectedRows>();
+      tensor = selected_rows->mutable_value();
+      tensor->Resize(framework::make_ddim(shape));
+      selected_rows->mutable_rows()->reserve(shape[0]);
+    } else {
+      PADDLE_THROW(platform::errors::InvalidArgument(
+          "Expected type of Output(out) in uniform_random_op must be "
+          "LoDTensor, "
+          "SelectedRows. But got unsupport type: %s.",
+          framework::ToTypeName(out_var->Type())));
+    }
+    T *data = tensor->mutable_data<T>(ctx.GetPlace());
+
+    int64_t size = tensor->numel();
+    std::uniform_real_distribution<T> dist(
+        static_cast<T>(ctx.Attr<float>("min")),
+        static_cast<T>(ctx.Attr<float>("max")));
+    unsigned int seed = static_cast<unsigned int>(ctx.Attr<int>("seed"));
+    // TODO(pangyoki): implement GetXPURandomEngine to set different seeds on
+    // corresponding XPU device.
+    auto engine = framework::GetCPURandomEngine(seed);
+
+    std::unique_ptr<T[]> data_cpu(new T[size]);
+    for (int64_t i = 0; i < size; ++i) {
+      data_cpu[i] = dist(*engine);
+    }
+
+    memory::Copy(BOOST_GET_CONST(platform::XPUPlace, ctx.GetPlace()), data,
+                 platform::CPUPlace(), reinterpret_cast<void *>(data_cpu.get()),
+                 size * sizeof(T));
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+REGISTER_OP_XPU_KERNEL(uniform_random,
+                       paddle::operators::XPUUniformRandomKernel<float>);
+
+#endif  // PADDLE_WITH_XPU
diff --git a/python/paddle/fluid/tests/unittests/xpu/test_uniform_random_op_xpu.py b/python/paddle/fluid/tests/unittests/xpu/test_uniform_random_op_xpu.py
new file mode 100644
index 00000000000..ab59fd26656
--- /dev/null
+++ b/python/paddle/fluid/tests/unittests/xpu/test_uniform_random_op_xpu.py
@@ -0,0 +1,51 @@
+#   Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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 __future__ import print_function
+
+import sys
+sys.path.append("..")
+import subprocess
+import unittest
+import numpy as np
+from op_test import OpTest
+import paddle
+import paddle.fluid.core as core
+from paddle.fluid.op import Operator
+import paddle.fluid as fluid
+from paddle.fluid import Program, program_guard
+from test_uniform_random_op import TestUniformRandomOp, TestUniformRandomOpSelectedRows
+
+paddle.enable_static()
+
+
+class TestXPUUniformRandomOp(TestUniformRandomOp):
+    def test_check_output(self):
+        if paddle.is_compiled_with_xpu():
+            place = paddle.XPUPlace(0)
+            outs = self.calc_output(place)
+            outs = [np.array(out) for out in outs]
+            outs.sort(key=len)
+            self.verify_output(outs)
+
+
+class TestXPUUniformRandomOpSelectedRows(TestUniformRandomOpSelectedRows):
+    def test_check_output(self):
+        if paddle.is_compiled_with_xpu():
+            place = paddle.XPUPlace(0)
+            self.check_with_place(place)
+
+
+if __name__ == "__main__":
+    unittest.main()
-- 
GitLab