diff --git a/paddle/fluid/operators/reshape2_op_npu.cc b/paddle/fluid/operators/reshape2_op_npu.cc new file mode 100644 index 0000000000000..7ca85abcf7afc --- /dev/null +++ b/paddle/fluid/operators/reshape2_op_npu.cc @@ -0,0 +1,87 @@ +/* Copyright (c) 2021 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. */ + +#include +#include + +#include "paddle/fluid/operators/npu_op_runner.h" + +namespace paddle { +namespace operators { + +template +class Reshape2NPUKernel : public framework::OpKernel { + public: + void Compute(const framework::ExecutionContext& ctx) const override { + auto* x = ctx.Input("X"); + auto* shape = ctx.Attr>> ("shape"); + auto* out = ctx.Output("Out"); + auto org_shape = framework::vectorize(x->dims()); + // reshape + int64_t shape_all = 1; + int64_t org_shape_all = 1; + int index = -1; + for (int i = 0; i < shape.size(); i++) { + if (shape[i] == 0) { + shape[i] = org_shape[i]; + } + if (shape[i] == -1) { + index = i; + } else { + shape_all *= shape[i]; + } + org_shape_all *= org_shape[i]; + } + + if (index >= 0) { + shape[index] = org_shape_all / shape_all; + } + out.Resize(framework::make_ddim(shape)); + out->mutable_data(ctx.GetPlace(), x->type()); + framework::TensorCopy( + *x, ctx.GetPlace(), + ctx.template device_context(), out); + out.Resize(framework::make_ddim(shape)); + } +}; + +template +class Reshape2GradNPUKernel : public framework::OpKernel { + public: + void Compute(const framework::ExecutionContext& ctx) const override { + auto* d_x = ctx.Output(framework::GradVarName("X")); + auto* d_out = ctx.Input(framework::GradVarName("Out")); + auto in_dims = d_x->dims(); + + d_x->mutable_data(ctx.GetPlace(), d_out->type()); + framework::TensorCopy( + *d_out, ctx.GetPlace(), + ctx.template device_context(), d_x); + d_x->Resize(in_dims); + } +}; +} // namespace operators +} // namespace paddle + +namespace ops = paddle::operators; + +REGISTER_OP_NPU_KERNEL( + reshpe2, ops::Reshape2NPUKernel, + ops::Reshape2NPUKernel); +REGISTER_OP_NPU_KERNEL( + reshpe2_grad, + ops::Reshape2GradNPUKernel, + ops::Reshape2GradNPUKernel); diff --git a/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py new file mode 100644 index 0000000000000..fe6eb6b5189cd --- /dev/null +++ b/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py @@ -0,0 +1,141 @@ +# Copyright (c) 2021 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 numpy as np +import unittest +import sys +sys.path.append("..") +from op_test import OpTest +import paddle +import paddle.fluid as fluid + +paddle.enable_static() +SEED = 2021 + + +@unittest.skipIf(not paddle.is_compiled_with_npu(), + "core is not compiled with NPU") +class TestReshape2(OpTest): + def setUp(self): + self.set_npu() + self.op_type = "reshape2" + self.place = paddle.NPUPlace(0) + + self.init_data() + self.inputs = {"X": np.random.random(self.ori_shape).astype("float32")} + self.attrs = {"shape": self.new_shape} + self.outputs = { + "Out": self.inputs["X"].reshape(self.infered_shape), + 'XShape': np.random.random(self.ori_shape).astype("float32") + } + + def set_npu(self): + self.__class__.use_npu = True + + def init_data(self): + self.ori_shape = (2, 60) + self.new_shape = (12, 10) + self.infered_shape = (12, 10) + + def test_check_output(self): + self.check_output( + self.place, check_dygraph=False, no_check_set=['XShape']) + + +class TestReshape2_case2(TestReshape2): + def init_data(self): + self.ori_shape = (2, 60) + self.new_shape = (-1, 10) + self.infered_shape = (12, 10) + + +class TestReshape2_case3(TestReshape2): + def init_data(self): + self.ori_shape = (2, 5, 6) + self.new_shape = (-1, 0, 3) + self.infered_shape = (4, 5, 3) + + + # TODO(ascendrc): Add grad test + # def test_check_grad(self): + # if self.dtype == np.float16: + # return + # self.check_grad(['X'], 'Out') + # +@unittest.skipIf(not paddle.is_compiled_with_npu(), + "core is not compiled with NPU") +class TestReshapeNet(unittest.TestCase): + def _test(self, run_npu=True): + main_prog = paddle.static.Program() + startup_prog = paddle.static.Program() + main_prog.random_seed = SEED + startup_prog.random_seed = SEED + np.random.seed(SEED) + + a_np = np.random.random(size=(32, 32)).astype('float32') + b_np = np.random.random(size=(32, 32)).astype('float32') + label_np = np.random.randint(2, size=(32, 1)).astype('int64') + + with paddle.static.program_guard(main_prog, startup_prog): + a = paddle.static.data(name="a", shape=[32, 32], dtype='float32') + b = paddle.static.data(name="b", shape=[32, 32], dtype='float32') + label = paddle.static.data( + name="label", shape=[32, 1], dtype='int64') + + sum = paddle.add(a, b) + z = paddle.reshape(sum, shape=[32, 32]) + + fc_1 = fluid.layers.fc(input=z, size=128) + prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax') + + cost = fluid.layers.cross_entropy(input=prediction, label=label) + loss = fluid.layers.reduce_mean(cost) + sgd = fluid.optimizer.SGD(learning_rate=0.01) + sgd.minimize(loss) + + if run_npu: + place = paddle.NPUPlace(0) + else: + place = paddle.CPUPlace() + + exe = paddle.static.Executor(place) + exe.run(startup_prog) + + print("Start run on {}".format(place)) + for epoch in range(100): + + pred_res, loss_res = exe.run( + main_prog, + feed={"a": a_np, + "b": b_np, + "label": label_np}, + fetch_list=[prediction, loss]) + if epoch % 10 == 0: + print("Epoch {} | Prediction[0]: {}, Loss: {}".format( + epoch, pred_res[0], loss_res)) + + return pred_res, loss_res + + def test_npu(self): + cpu_pred, cpu_loss = self._test(False) + npu_pred, npu_loss = self._test(True) + + self.assertTrue(np.allclose(npu_pred, cpu_pred)) + self.assertTrue(np.allclose(npu_loss, cpu_loss)) + + +if __name__ == '__main__': + unittest.main()