remove fluid.layer.gather_tree (#48480)

python/paddle/fluid/layers/nn.py

@@ -128,7 +128,6 @@ __all__ = [
     'shard_index',
     'hard_swish',
     'mish',
-    'gather_tree',
     'uniform_random',
     'unbind',
 ]
@@ -7928,70 +7927,6 @@ def mish(x, threshold=20, name=None):
     return out
 
 
-def gather_tree(ids, parents):
-    r"""
-    To be used after beam search. After beam search, we get selected ids at
-    each time step and the corresponding parents in the search tree. Both ids
-    and parents have the layout :attr:`[max_time, batch_size, beam_size]`. Then
-    :attr:`gather_tree` is used to backtrace from the last time step and
-    generate the full sequences by collecting selected ids.
-
-    Here is an example:
-
-    .. code-block:: text
-
-            Given:
-                ids = [[[2 2]
-                        [6 1]]
-                       [[3 9]
-                        [6 1]]
-                       [[0 1]
-                        [9 0]]]
-                parents = [[[0 0]
-                            [1 1]]
-                           [[1 0]
-                            [1 0]]
-                           [[0 0]
-                            [0 1]]]
-
-            Then:
-                gather_tree(ids, parents)
-                         = [[[2 2]
-                             [1 6]]
-                            [[3 3]
-                             [6 1]]
-                            [[0 1]
-                             [9 0]]]
-
-    Args:
-        ids(Tensor): A Tensor with shape :attr:`[length, batch_size, beam_size]`
-            and data type :attr:`int32` or :attr:`int64`. It contains the selected
-            ids of all time steps.
-        parents(Tensor): A Tensor with the same shape and data type as :attr:`ids`,
-            It contains the parents corresponding to selected ids when searching
-            among beams.
-
-    Returns:
-            A Tensor with the same shape and data type as :attr:`ids`. \
-            It contains the full sequences. The sequences are collected from \
-            :attr:`ids` by backtracing according to :attr:`parents`.
-
-    Examples:
-        .. code-block:: python
-
-            import paddle
-
-            ids = paddle.to_tensor([[[2, 2], [6, 1]], [[3, 9], [6, 1]], [[0, 1], [9, 0]]])
-
-            parents = paddle.to_tensor([[[0, 0], [1, 1]], [[1, 0], [1, 0]], [[0, 0], [0, 1]]])
-
-            final_sequences = paddle.nn.functional.gather_tree(ids, parents)
-            # [[[2, 2], [1, 6]], [[3, 3], [6, 1]], [[0, 1], [9, 0]]]
-
-    """
-    return paddle.nn.functional.gather_tree(ids, parents)
-
-
 @deprecated(since="2.0.0", update_to="paddle.uniform")
 @templatedoc()
 def uniform_random(

python/paddle/fluid/layers/rnn.py

@@ -1427,7 +1427,7 @@ class BeamSearchDecoder(Decoder):
                 `[time_step, batch_size, beam_size]`. `final_states` is the same \
                 as the input argument `final_states`.
         """
-        predicted_ids = nn.gather_tree(
+        predicted_ids = paddle.nn.functional.gather_tree(
             outputs.predicted_ids, outputs.parent_ids
         )
         # TODO: use FinalBeamSearchDecoderOutput as output

python/paddle/fluid/tests/unittests/dygraph_to_static/seq2seq_dygraph_model.py

@@ -502,7 +502,9 @@ class BaseModel(fluid.dygraph.Layer):
 
         predicted_ids = paddle.stack(predicted_ids)
         parent_ids = paddle.stack(parent_ids)
-        predicted_ids = fluid.layers.gather_tree(predicted_ids, parent_ids)
+        predicted_ids = paddle.nn.functional.gather_tree(
+            predicted_ids, parent_ids
+        )
         predicted_ids = self._transpose_batch_time(predicted_ids)
         return predicted_ids
 

python/paddle/fluid/tests/unittests/dygraph_to_static/transformer_dygraph_model.py

@@ -884,7 +884,7 @@ class Transformer(Layer):
         predict_ids = paddle.stack(predict_ids, axis=0)
         parent_ids = paddle.stack(parent_ids, axis=0)
         finished_seq = paddle.transpose(
-            layers.gather_tree(predict_ids, parent_ids), [1, 2, 0]
+            paddle.nn.functional.gather_tree(predict_ids, parent_ids), [1, 2, 0]
         )
         finished_scores = topk_scores
 

python/paddle/fluid/tests/unittests/test_gather_tree_op.py

@@ -67,7 +67,7 @@ class TestGatherTreeOpAPI(unittest.TestCase):
             dtype='int64',
             append_batch_size=False,
         )
-        final_sequences = fluid.layers.gather_tree(ids, parents)
+        final_sequences = paddle.nn.functional.gather_tree(ids, parents)
         paddle.disable_static()
 
     def test_case2(self):
@@ -100,14 +100,14 @@ class TestGatherTreeOpError(unittest.TestCase):
             def test_Variable_ids():
                 # the input type must be Variable
                 np_ids = np.random.random((5, 2, 2), dtype='int64')
-                fluid.layers.gather_tree(np_ids, parents)
+                paddle.nn.functional.gather_tree(np_ids, parents)
 
             self.assertRaises(TypeError, test_Variable_ids)
 
             def test_Variable_parents():
                 # the input type must be Variable
                 np_parents = np.random.random((5, 2, 2), dtype='int64')
-                fluid.layers.gather_tree(ids, np_parents)
+                paddle.nn.functional.gather_tree(ids, np_parents)
 
             self.assertRaises(TypeError, test_Variable_parents)
 
@@ -119,7 +119,7 @@ class TestGatherTreeOpError(unittest.TestCase):
                     dtype='float32',
                     append_batch_size=False,
                 )
-                fluid.layers.gather_tree(bad_ids, parents)
+                paddle.nn.functional.gather_tree(bad_ids, parents)
 
             self.assertRaises(TypeError, test_type_ids)
 
@@ -131,7 +131,7 @@ class TestGatherTreeOpError(unittest.TestCase):
                     dtype='float32',
                     append_batch_size=False,
                 )
-                fluid.layers.gather_tree(ids, bad_parents)
+                paddle.nn.functional.gather_tree(ids, bad_parents)
 
             self.assertRaises(TypeError, test_type_parents)