Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add_dilation

paddle/operators/beam_search_decode_op.cc

@@ -27,6 +27,7 @@ class BeamSearchDecodeOp : public framework::OperatorBase {
   void Run(const framework::Scope& scope,
            const platform::DeviceContext& dev_ctx) const override {
     framework::ExecutionContext ctx(*this, scope, dev_ctx);
+
     const LoDTensorArray* ids = ctx.Input<LoDTensorArray>("Ids");
     const LoDTensorArray* scores = ctx.Input<LoDTensorArray>("Scores");
     const size_t step_num = ids->size();

paddle/operators/l1_norm_op.h

@@ -29,7 +29,7 @@ class L1NormKernel : public framework::OpKernel<T> {
     Out->mutable_data<T>(context.GetPlace());
 
     auto x = framework::EigenVector<T>::Flatten(*X);
-    auto out = framework::EigenVector<T>::Flatten(*Out);
+    auto out = framework::EigenScalar<T>::From(*Out);
     auto place = context.GetEigenDevice<Place>();
 
     out.device(place) = x.abs().sum();

paddle/operators/squared_l2_norm_op.h

@@ -29,7 +29,7 @@ class SquaredL2NormKernel : public framework::OpKernel<T> {
     Out->mutable_data<T>(context.GetPlace());
 
     auto x = framework::EigenVector<T>::Flatten(*X);
-    auto out = framework::EigenVector<T>::Flatten(*Out);
+    auto out = framework::EigenScalar<T>::From(*Out);
     auto place = context.GetEigenDevice<Place>();
 
     out.device(place) = x.square().sum();

paddle/pybind/pybind.cc

@@ -42,6 +42,9 @@ limitations under the License. */
 #include "paddle/platform/gpu_info.h"
 #endif
 
+// disable auto conversion to list in Python
+PYBIND11_MAKE_OPAQUE(paddle::framework::LoDTensorArray);
+
 namespace paddle {
 namespace pybind {
 static size_t UniqueIntegerGenerator(const std::string &prefix) {

python/paddle/v2/framework/layers.py

@@ -839,6 +839,23 @@ def batch_norm(input,
     return helper.append_activation(batch_norm_out)
 
 
+def beam_search_decode(ids, scores, main_program=None, startup_program=None):
+    helper = LayerHelper('beam_search_decode', **locals())
+    sentence_ids = helper.create_tmp_variable(dtype=ids.data_type)
+    sentence_scores = helper.create_tmp_variable(dtype=ids.data_type)
+
+    helper.append_op(
+        type="beam_search_decode",
+        inputs={"Ids": ids,
+                "Scores": scores},
+        outputs={
+            "SentenceIds": sentence_ids,
+            "SentenceScores": sentence_scores
+        })
+
+    return sentence_ids, sentence_scores
+
+
 class BlockGuard(object):
     """
     BlockGuard class.

python/paddle/v2/framework/tests/test_beam_search_decode_op.py

+import unittest
+
+import numpy as np
+import paddle.v2.framework.core as core
+from paddle.v2.framework.op import Operator
+
+
+class TestBeamSearchDecodeOp(unittest.TestCase):
+    def setUp(self):
+        self.scope = core.Scope()
+        self.cpu_place = core.CPUPlace()
+
+    def append_lod_tensor(self, tensor_array, lod, data):
+        lod_tensor = core.LoDTensor()
+        lod_tensor.set_lod(lod)
+        lod_tensor.set(data, self.cpu_place)
+        tensor_array.append(lod_tensor)
+
+    def test_get_set(self):
+        ids = self.scope.var("ids").get_lod_tensor_array()
+        self.append_lod_tensor(
+            ids, [[0, 3, 6], [0, 1, 2, 3, 4, 5, 6]],
+            np.array(
+                [1, 2, 3, 4, 5, 6], dtype="int64"))
+        self.append_lod_tensor(
+            ids, [[0, 3, 6], [0, 1, 1, 3, 5, 5, 6]],
+            np.array(
+                [0, 1, 2, 3, 4, 5], dtype="int64"))
+        self.append_lod_tensor(
+            ids, [[0, 3, 6], [0, 0, 1, 2, 3, 4, 5]],
+            np.array(
+                [0, 1, 2, 3, 4], dtype="int64"))
+
+        scores = self.scope.var("scores").get_lod_tensor_array()
+        self.append_lod_tensor(
+            scores, [[0, 3, 6], [0, 1, 2, 3, 4, 5, 6]],
+            np.array(
+                [1, 2, 3, 4, 5, 6], dtype="float32"))
+        self.append_lod_tensor(
+            scores, [[0, 3, 6], [0, 1, 1, 3, 5, 5, 6]],
+            np.array(
+                [0, 1, 2, 3, 4, 5], dtype="float32"))
+        self.append_lod_tensor(
+            scores, [[0, 3, 6], [0, 0, 1, 2, 3, 4, 5]],
+            np.array(
+                [0, 1, 2, 3, 4], dtype="float32"))
+
+        sentence_ids = self.scope.var("sentence_ids").get_tensor()
+        sentence_scores = self.scope.var("sentence_scores").get_tensor()
+
+        beam_search_decode_op = Operator(
+            "beam_search_decode",
+            # inputs
+            Ids="ids",
+            Scores="scores",
+            # outputs
+            SentenceIds="sentence_ids",
+            SentenceScores="sentence_scores")
+
+        ctx = core.DeviceContext.create(self.cpu_place)
+        beam_search_decode_op.run(self.scope, ctx)
+
+        expected_lod = [[0, 4, 8], [0, 1, 3, 6, 9, 10, 13, 16, 19]]
+        self.assertEqual(sentence_ids.lod(), expected_lod)
+        self.assertEqual(sentence_scores.lod(), expected_lod)
+
+        expected_data = np.array(
+            [2, 1, 0, 3, 1, 0, 3, 2, 1, 5, 4, 3, 2, 4, 4, 3, 6, 5, 4], "int64")
+        self.assertTrue(np.array_equal(np.array(sentence_ids), expected_data))
+        self.assertTrue(
+            np.array_equal(np.array(sentence_scores), expected_data))
+
+
+if __name__ == '__main__':
+    unittest.main()