diff --git a/paddle/operators/sequence_pool_op.cc b/paddle/operators/sequence_pool_op.cc index e3f5d509a85537669237b8fd0ed44efe8abb6874..6d600c27271c660f0cf933e8bd05455df61740ec 100644 --- a/paddle/operators/sequence_pool_op.cc +++ b/paddle/operators/sequence_pool_op.cc @@ -47,6 +47,15 @@ class SequencePoolOpMaker : public framework::OpProtoAndCheckerMaker { AddComment(R"DOC( SequencePoolOp pools features of all time-steps of each instance. + It supports six pooling strategy: + - AVERAGE: Out[i] = average_{for each instance in i-th sequence}{X[i]} + - SUM: Out[i] = sum_{for each instance in i-th sequence}{X[i]} + - SQRT: Out[i] = sum_{for each instance in i-th sequence}{X[i]} + / sqrt(i-th sequence length) + - LAST: Out[i] = last instance in i-th sequence X[i] + - FIRST: Out[i] = first instance in i-th sequence X[i] + - MAX: Out[i] = max_{for each instance in i-th sequence}{X[i]} + For a mini-batch of 3 variable-length sentences, containing 2, 3, and 2 time-steps: Assume X is a [7,M,N] LoDTensor, and X->lod()[0] = [0, 2, 5, 7], 7=2+3+2. diff --git a/python/paddle/v2/framework/tests/test_seq_pool.py b/python/paddle/v2/framework/tests/test_seq_pool.py index 58a555f7737a540b338b79ae2d83145cc9b568e3..591494e83c3a48094f099574272e4cff7a46ccf2 100644 --- a/python/paddle/v2/framework/tests/test_seq_pool.py +++ b/python/paddle/v2/framework/tests/test_seq_pool.py @@ -16,24 +16,23 @@ class TestSeqAvgPool(OpTest): def set_data(self): self.op_type = 'sequence_pool' # one level, batch size is 4 - x = np.random.uniform(0.1, 1, [11, 2]).astype('float32') + x = np.random.uniform(0.1, 1, [11, 23]).astype('float32') lod = [[0, 4, 5, 8, 11]] self.inputs = {'X': (x, lod)} - out = np.zeros((4, 2)).astype('float32') + out = np.zeros((4, 23)).astype('float32') self.outputs = {'Out': out} + return x, lod, out - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.AVERAGE} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] out[i] = sub_x.mean(axis=0) def setUp(self): - self.set_data() - self.compute() + x, lod, out = self.set_data() + self.compute(x, lod, out) def test_check_output(self): self.check_output() @@ -52,41 +51,34 @@ class TestSeqAvgPool2D(TestSeqAvgPool): out = np.zeros((4, 3, 17)).astype('float32') self.outputs = {'Out': out} + return x, lod, out - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.AVERAGE} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) out[i] = np.reshape(sub_x.mean(axis=0), (3, 17)) class TestSeqSumPool(TestSeqAvgPool): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.SUM} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] out[i] = sub_x.sum(axis=0) class TestSeqSumPool2D(TestSeqAvgPool2D): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.SUM} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) out[i] = np.reshape(sub_x.sum(axis=0), (3, 17)) class TestSeqSqrtPool(TestSeqAvgPool): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.SQRT} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] len = lod[0][i + 1] - lod[0][i] @@ -94,10 +86,8 @@ class TestSeqSqrtPool(TestSeqAvgPool): class TestSeqSqrtPool2D(TestSeqAvgPool2D): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.SQRT} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) len = lod[0][i + 1] - lod[0][i] @@ -108,20 +98,16 @@ class TestSeqSqrtPool2D(TestSeqAvgPool2D): class TestSeqMaxPool(TestSeqAvgPool): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.MAX} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] out[i] = np.amax(sub_x, axis=0) class TestSeqMaxPool2D(TestSeqAvgPool2D): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.MAX} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) out[i] = np.reshape(np.amax(sub_x, axis=0), (3, 17)) @@ -132,40 +118,32 @@ class TestSeqMaxPool2D(TestSeqAvgPool2D): class TestSeqLastPool(TestSeqAvgPool): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.LAST} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] out[i] = sub_x[-1, :] class TestSeqLastPool2D(TestSeqAvgPool2D): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.LAST} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) out[i] = np.reshape(sub_x[-1, :], (3, 17)) class TestSeqFirstPool(TestSeqAvgPool): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.FIRST} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = x[lod[0][i]:lod[0][i + 1], :] out[i] = sub_x[0, :] class TestSeqFirstPool2D(TestSeqAvgPool2D): - def compute(self): + def compute(self, x, lod, out): self.attrs = {'strategy': SeqPoolType.FIRST} - x, lod = self.inputs['X'] - out = self.outputs['Out'] for i in range(4): sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17)) out[i] = np.reshape(sub_x[0, :], (3, 17))