Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2299
Star 20931
Fork 5422
P
Paddle
提交 54e07994 编写于 作者: Y Youwei Song 提交者： Tao Luo
浏览文件
操作

Dygraph Layer kwargs & param getter setter (#19901) 

* opt FC

* opt rest of dygraph.nn

* new param shape check and unittest

* add kwargs for Layer

* add new set_value api

* use property decorator

* update API.spec, test=develop

* use UserList, separate gettersetters, test=develop

* update test_custom_layer_with_kwargs, test=develop

* fix UserList compatibility, test=develop

* fix UserList compatibility, test=develop

* keep FC._w, test=develop

* add unittests, Conv3D bug fix, test=develop

* clean code, test=develop

* fix dygraph guard in unittest, test=develop

* add property setters, remove unused param in tracer, test=develop

* tracer none check, test=develop

* merge, test=develop

* refine, test=develop

* bug fix in  prelu and conv3d_transpose, test=develop

* rm __set__, test=develop

* set tensor value instead of assign op

* fix property setter call, test=develop

* fix api.spec, test=develop

* fix doc sample, test=develop
上级 9de67725
显示空白变更内容
内联 并排
Showing with 770 addition and 65 deletion
paddle/fluid/API.spec
浏览文件 @ 54e07994
......@@ -24,6 +24,7 @@ paddle.fluid.Variable.clear_gradient (ArgSpec(args=['self'], varargs=None, keywo
paddle.fluid.Variable.detach (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '0730b2d310b014d9b0a903b2034757d7'))
paddle.fluid.Variable.gradient (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '86b246bfaf20f3058e91927abbcf9fb9'))
paddle.fluid.Variable.numpy (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '7536e8feb56d827875943e7f01d406fc'))
paddle.fluid.Variable.set_value (ArgSpec(args=['self', 'value'], varargs=None, keywords=None, defaults=None), ('document', 'c424b9e763ff51c38a6917f98026fe7d'))
paddle.fluid.Variable.to_string (ArgSpec(args=['self', 'throw_on_error', 'with_details'], varargs=None, keywords=None, defaults=(False,)), ('document', '31f359a2c074f26dc0ffff296fc3983f'))
paddle.fluid.Executor ('paddle.fluid.executor.Executor', ('document', '34e8c1769313fbeff7817212dda6259e'))
paddle.fluid.Executor.__init__ (ArgSpec(args=['self', 'place'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
......@@ -575,7 +576,7 @@ paddle.fluid.dygraph.Layer.clear_gradients (ArgSpec(args=['self'], varargs=None,
paddle.fluid.dygraph.Layer.create_parameter (ArgSpec(args=['self', 'attr', 'shape', 'dtype', 'is_bias', 'default_initializer'], varargs=None, keywords=None, defaults=(False, None)), ('document', 'a6420ca1455366eaaf972191612de0b6'))
paddle.fluid.dygraph.Layer.create_variable (ArgSpec(args=['self', 'name', 'persistable', 'dtype', 'type'], varargs=None, keywords=None, defaults=(None, None, None, VarType.LOD_TENSOR)), ('document', '171cccfceba636d5bbf7bbae672945d8'))
paddle.fluid.dygraph.Layer.eval (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Layer.forward (ArgSpec(args=['self'], varargs='inputs', keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Layer.forward (ArgSpec(args=['self'], varargs='inputs', keywords='kwargs', defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Layer.full_name (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '23ce4f961f48ed0f79cadf93a3938ed2'))
paddle.fluid.dygraph.Layer.load_dict (ArgSpec(args=['self', 'stat_dict', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Layer.parameters (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '5aec25a854eb57abc798dccccbb507d5'))
......@@ -585,7 +586,7 @@ paddle.fluid.dygraph.Layer.train (ArgSpec(args=['self'], varargs=None, keywords=
paddle.fluid.dygraph.__impl__ (ArgSpec(args=['func'], varargs=None, keywords=None, defaults=()), ('document', '75d1d3afccc8b39cdebf05cb1f5969f9'))
paddle.fluid.dygraph.guard (ArgSpec(args=['place'], varargs=None, keywords=None, defaults=(None,)), ('document', '7071320ffe2eec9aacdae574951278c6'))
paddle.fluid.dygraph.to_variable (ArgSpec(args=['value', 'block', 'name'], varargs=None, keywords=None, defaults=(None, None)), ('document', '0e69fa3666f15dd01b6e3e270b9371cd'))
paddle.fluid.dygraph.Conv2D ('paddle.fluid.dygraph.nn.Conv2D', ('document', 'baafe7ae0d3a61ae79cf4c7443e2c37c'))
paddle.fluid.dygraph.Conv2D ('paddle.fluid.dygraph.nn.Conv2D', ('document', '0b6acb9cc7fbb4f5b129e1f6dd985581'))
paddle.fluid.dygraph.Conv2D.__init__ (ArgSpec(args=['self', 'name_scope', 'num_filters', 'filter_size', 'stride', 'padding', 'dilation', 'groups', 'param_attr', 'bias_attr', 'use_cudnn', 'act', 'dtype'], varargs=None, keywords=None, defaults=(1, 0, 1, None, None, None, True, None, 'float32')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv2D.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Conv2D.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -601,7 +602,7 @@ paddle.fluid.dygraph.Conv2D.parameters (ArgSpec(args=['self', 'include_sublayers
paddle.fluid.dygraph.Conv2D.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv2D.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Conv2D.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3D ('paddle.fluid.dygraph.nn.Conv3D', ('document', '8b756aaca65af9594cc574d9a5d2b055'))
paddle.fluid.dygraph.Conv3D ('paddle.fluid.dygraph.nn.Conv3D', ('document', '50412bd3fbf3557a8ef48e25c6517025'))
paddle.fluid.dygraph.Conv3D.__init__ (ArgSpec(args=['self', 'name_scope', 'num_filters', 'filter_size', 'stride', 'padding', 'dilation', 'groups', 'param_attr', 'bias_attr', 'use_cudnn', 'act'], varargs=None, keywords=None, defaults=(1, 0, 1, None, None, None, True, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3D.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Conv3D.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -617,7 +618,7 @@ paddle.fluid.dygraph.Conv3D.parameters (ArgSpec(args=['self', 'include_sublayers
paddle.fluid.dygraph.Conv3D.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3D.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Conv3D.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Pool2D ('paddle.fluid.dygraph.nn.Pool2D', ('document', 'e9331666e47a38586c8809a23cbaf7de'))
paddle.fluid.dygraph.Pool2D ('paddle.fluid.dygraph.nn.Pool2D', ('document', '50e6fd200e42859daf2924ecb0561ada'))
paddle.fluid.dygraph.Pool2D.__init__ (ArgSpec(args=['self', 'name_scope', 'pool_size', 'pool_type', 'pool_stride', 'pool_padding', 'global_pooling', 'use_cudnn', 'ceil_mode', 'exclusive', 'dtype'], varargs=None, keywords=None, defaults=(-1, 'max', 1, 0, False, True, False, True, VarType.FP32)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Pool2D.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Pool2D.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -633,7 +634,7 @@ paddle.fluid.dygraph.Pool2D.parameters (ArgSpec(args=['self', 'include_sublayers
paddle.fluid.dygraph.Pool2D.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Pool2D.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Pool2D.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.FC ('paddle.fluid.dygraph.nn.FC', ('document', '1d64242f03f2aca2307e94590b552430'))
paddle.fluid.dygraph.FC ('paddle.fluid.dygraph.nn.FC', ('document', '2f73ae00e57c67454c6aa7e911d9bfd6'))
paddle.fluid.dygraph.FC.__init__ (ArgSpec(args=['self', 'name_scope', 'size', 'num_flatten_dims', 'param_attr', 'bias_attr', 'act', 'is_test', 'dtype'], varargs=None, keywords=None, defaults=(1, None, None, None, False, 'float32')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.FC.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.FC.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -649,7 +650,7 @@ paddle.fluid.dygraph.FC.parameters (ArgSpec(args=['self', 'include_sublayers'],
paddle.fluid.dygraph.FC.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.FC.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.FC.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BatchNorm ('paddle.fluid.dygraph.nn.BatchNorm', ('document', '0b609e10e4d417c91d346f887d16771c'))
paddle.fluid.dygraph.BatchNorm ('paddle.fluid.dygraph.nn.BatchNorm', ('document', '390fb9b986423ec6680731ffc7cf24ab'))
paddle.fluid.dygraph.BatchNorm.__init__ (ArgSpec(args=['self', 'name_scope', 'num_channels', 'act', 'is_test', 'momentum', 'epsilon', 'param_attr', 'bias_attr', 'dtype', 'data_layout', 'in_place', 'moving_mean_name', 'moving_variance_name', 'do_model_average_for_mean_and_var', 'fuse_with_relu', 'use_global_stats', 'trainable_statistics'], varargs=None, keywords=None, defaults=(None, False, 0.9, 1e-05, None, None, 'float32', 'NCHW', False, None, None, False, False, False, False)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BatchNorm.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.BatchNorm.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -665,7 +666,7 @@ paddle.fluid.dygraph.BatchNorm.parameters (ArgSpec(args=['self', 'include_sublay
paddle.fluid.dygraph.BatchNorm.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BatchNorm.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.BatchNorm.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Embedding ('paddle.fluid.dygraph.nn.Embedding', ('document', 'ecf8dc4149f005cd30eddc0dd343454f'))
paddle.fluid.dygraph.Embedding ('paddle.fluid.dygraph.nn.Embedding', ('document', 'b1b1ed9dc2125c3e16ee08113605fcb4'))
paddle.fluid.dygraph.Embedding.__init__ (ArgSpec(args=['self', 'name_scope', 'size', 'is_sparse', 'is_distributed', 'padding_idx', 'param_attr', 'dtype'], varargs=None, keywords=None, defaults=(False, False, None, None, 'float32')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Embedding.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Embedding.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -681,7 +682,7 @@ paddle.fluid.dygraph.Embedding.parameters (ArgSpec(args=['self', 'include_sublay
paddle.fluid.dygraph.Embedding.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Embedding.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Embedding.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GRUUnit ('paddle.fluid.dygraph.nn.GRUUnit', ('document', '5308e42b6a6db4681ce5ee9e94983986'))
paddle.fluid.dygraph.GRUUnit ('paddle.fluid.dygraph.nn.GRUUnit', ('document', '389e860e455b67aab1f4d472ac9d7e49'))
paddle.fluid.dygraph.GRUUnit.__init__ (ArgSpec(args=['self', 'name_scope', 'size', 'param_attr', 'bias_attr', 'activation', 'gate_activation', 'origin_mode', 'dtype'], varargs=None, keywords=None, defaults=(None, None, 'tanh', 'sigmoid', False, 'float32')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GRUUnit.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.GRUUnit.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -697,7 +698,7 @@ paddle.fluid.dygraph.GRUUnit.parameters (ArgSpec(args=['self', 'include_sublayer
paddle.fluid.dygraph.GRUUnit.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GRUUnit.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.GRUUnit.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.LayerNorm ('paddle.fluid.dygraph.nn.LayerNorm', ('document', 'b44f5d3d10386c460094e21f24ff272b'))
paddle.fluid.dygraph.LayerNorm ('paddle.fluid.dygraph.nn.LayerNorm', ('document', '8bc39f59fe2d3713bc143fdf1222a63b'))
paddle.fluid.dygraph.LayerNorm.__init__ (ArgSpec(args=['self', 'name_scope', 'scale', 'shift', 'begin_norm_axis', 'epsilon', 'param_attr', 'bias_attr', 'act'], varargs=None, keywords=None, defaults=(True, True, 1, 1e-05, None, None, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.LayerNorm.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.LayerNorm.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -713,7 +714,7 @@ paddle.fluid.dygraph.LayerNorm.parameters (ArgSpec(args=['self', 'include_sublay
paddle.fluid.dygraph.LayerNorm.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.LayerNorm.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.LayerNorm.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.NCE ('paddle.fluid.dygraph.nn.NCE', ('document', '2d579e8d9ce31bb29e079e5f6108fc73'))
paddle.fluid.dygraph.NCE ('paddle.fluid.dygraph.nn.NCE', ('document', '993aeea9be436e9c709a758795cb23e9'))
paddle.fluid.dygraph.NCE.__init__ (ArgSpec(args=['self', 'name_scope', 'num_total_classes', 'sample_weight', 'param_attr', 'bias_attr', 'num_neg_samples', 'sampler', 'custom_dist', 'seed', 'is_sparse'], varargs=None, keywords=None, defaults=(None, None, None, None, 'uniform', None, 0, False)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.NCE.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.NCE.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -729,7 +730,7 @@ paddle.fluid.dygraph.NCE.parameters (ArgSpec(args=['self', 'include_sublayers'],
paddle.fluid.dygraph.NCE.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.NCE.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.NCE.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.PRelu ('paddle.fluid.dygraph.nn.PRelu', ('document', 'd395ed163b4cf86e7207968f27bc1c11'))
paddle.fluid.dygraph.PRelu ('paddle.fluid.dygraph.nn.PRelu', ('document', 'da956af1676b08bf15553751a3643b55'))
paddle.fluid.dygraph.PRelu.__init__ (ArgSpec(args=['self', 'name_scope', 'mode', 'param_attr'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.PRelu.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.PRelu.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -745,7 +746,7 @@ paddle.fluid.dygraph.PRelu.parameters (ArgSpec(args=['self', 'include_sublayers'
paddle.fluid.dygraph.PRelu.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.PRelu.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.PRelu.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BilinearTensorProduct ('paddle.fluid.dygraph.nn.BilinearTensorProduct', ('document', '310140d784933928a27db9a7af4761e8'))
paddle.fluid.dygraph.BilinearTensorProduct ('paddle.fluid.dygraph.nn.BilinearTensorProduct', ('document', 'be70d0f6d43729d9cb80c9a34ed5f26b'))
paddle.fluid.dygraph.BilinearTensorProduct.__init__ (ArgSpec(args=['self', 'name_scope', 'size', 'name', 'act', 'param_attr', 'bias_attr'], varargs=None, keywords=None, defaults=(None, None, None, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BilinearTensorProduct.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.BilinearTensorProduct.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -761,7 +762,7 @@ paddle.fluid.dygraph.BilinearTensorProduct.parameters (ArgSpec(args=['self', 'in
paddle.fluid.dygraph.BilinearTensorProduct.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.BilinearTensorProduct.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.BilinearTensorProduct.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv2DTranspose ('paddle.fluid.dygraph.nn.Conv2DTranspose', ('document', '918fa8ad8a62ff424c842fb8a840bf7a'))
paddle.fluid.dygraph.Conv2DTranspose ('paddle.fluid.dygraph.nn.Conv2DTranspose', ('document', 'cf23c905abc00b07603dfa71a432d6f7'))
paddle.fluid.dygraph.Conv2DTranspose.__init__ (ArgSpec(args=['self', 'name_scope', 'num_filters', 'output_size', 'filter_size', 'padding', 'stride', 'dilation', 'groups', 'param_attr', 'bias_attr', 'use_cudnn', 'act'], varargs=None, keywords=None, defaults=(None, None, 0, 1, 1, None, None, None, True, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv2DTranspose.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Conv2DTranspose.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -777,7 +778,7 @@ paddle.fluid.dygraph.Conv2DTranspose.parameters (ArgSpec(args=['self', 'include_
paddle.fluid.dygraph.Conv2DTranspose.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv2DTranspose.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Conv2DTranspose.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3DTranspose ('paddle.fluid.dygraph.nn.Conv3DTranspose', ('document', 'cd99906d0813609ddea3fb6a2ac900dc'))
paddle.fluid.dygraph.Conv3DTranspose ('paddle.fluid.dygraph.nn.Conv3DTranspose', ('document', '91ba132bc690eaf76eabdbde8f87e4a0'))
paddle.fluid.dygraph.Conv3DTranspose.__init__ (ArgSpec(args=['self', 'name_scope', 'num_filters', 'output_size', 'filter_size', 'padding', 'stride', 'dilation', 'groups', 'param_attr', 'bias_attr', 'use_cudnn', 'act', 'name'], varargs=None, keywords=None, defaults=(None, None, 0, 1, 1, None, None, None, True, None, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3DTranspose.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.Conv3DTranspose.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -793,7 +794,7 @@ paddle.fluid.dygraph.Conv3DTranspose.parameters (ArgSpec(args=['self', 'include_
paddle.fluid.dygraph.Conv3DTranspose.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.Conv3DTranspose.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.Conv3DTranspose.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GroupNorm ('paddle.fluid.dygraph.nn.GroupNorm', ('document', '4d65fc6b00970e3b5c5dd0abeacd47cb'))
paddle.fluid.dygraph.GroupNorm ('paddle.fluid.dygraph.nn.GroupNorm', ('document', '72c125b07bdd1e612607dc77039b2722'))
paddle.fluid.dygraph.GroupNorm.__init__ (ArgSpec(args=['self', 'name_scope', 'groups', 'epsilon', 'param_attr', 'bias_attr', 'act', 'data_layout'], varargs=None, keywords=None, defaults=(1e-05, None, None, None, 'NCHW')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GroupNorm.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.GroupNorm.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -809,7 +810,7 @@ paddle.fluid.dygraph.GroupNorm.parameters (ArgSpec(args=['self', 'include_sublay
paddle.fluid.dygraph.GroupNorm.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.GroupNorm.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.GroupNorm.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.SpectralNorm ('paddle.fluid.dygraph.nn.SpectralNorm', ('document', 'f400a29393aa95fff829b4a6111e2952'))
paddle.fluid.dygraph.SpectralNorm ('paddle.fluid.dygraph.nn.SpectralNorm', ('document', '8f5cfbc431a8b4b44b605cde8b0381ef'))
paddle.fluid.dygraph.SpectralNorm.__init__ (ArgSpec(args=['self', 'name_scope', 'dim', 'power_iters', 'eps', 'name'], varargs=None, keywords=None, defaults=(0, 1, 1e-12, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.SpectralNorm.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.SpectralNorm.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......@@ -825,7 +826,7 @@ paddle.fluid.dygraph.SpectralNorm.parameters (ArgSpec(args=['self', 'include_sub
paddle.fluid.dygraph.SpectralNorm.state_dict (ArgSpec(args=['self', 'destination', 'include_sublayers'], varargs=None, keywords=None, defaults=(None, True)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.SpectralNorm.sublayers (ArgSpec(args=['self', 'include_sublayers'], varargs=None, keywords=None, defaults=(True,)), ('document', '00a881005ecbc96578faf94513bf0d62'))
paddle.fluid.dygraph.SpectralNorm.train (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.TreeConv ('paddle.fluid.dygraph.nn.TreeConv', ('document', '1e3104dea2482f6b79cf7a7ac9a343ab'))
paddle.fluid.dygraph.TreeConv ('paddle.fluid.dygraph.nn.TreeConv', ('document', '6e175a7bf2a43ae6c0f3a8a54bd69afe'))
paddle.fluid.dygraph.TreeConv.__init__ (ArgSpec(args=['self', 'name_scope', 'output_size', 'num_filters', 'max_depth', 'act', 'param_attr', 'bias_attr', 'name'], varargs=None, keywords=None, defaults=(1, 2, 'tanh', None, None, None)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.dygraph.TreeConv.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))
paddle.fluid.dygraph.TreeConv.add_sublayer (ArgSpec(args=['self', 'name', 'sublayer'], varargs=None, keywords=None, defaults=None), ('document', '839ff3c0534677ba6ad8735c3fd4e995'))
......
python/paddle/fluid/dygraph/layers.py
浏览文件 @ 54e07994
......@@ -150,20 +150,20 @@ class Layer(core.Layer):
if p.trainable:
p.clear_gradient()
def _build_once(self, *args):
def _build_once(self, *args, **kwargs):
pass
def __call__(self, *inputs):
def __call__(self, *inputs, **kwargs):
if not self._built:
self._build_once(*inputs)
self._build_once(*inputs, **kwargs)
if parallel_helper._is_data_parallel_mode():
parallel_helper._broadcast_parameters(self._parameters.values())
outputs = self.forward(*inputs)
outputs = self.forward(*inputs, **kwargs)
self._built = True
return outputs
def forward(self, *inputs):
def forward(self, *inputs, **kwargs):
raise NotImplementedError
def backward(self, *inputs):
......@@ -216,6 +216,8 @@ class Layer(core.Layer):
return object.__getattribute__(self, name)
def __setattr__(self, name, value):
if isinstance(getattr(type(self), name, None), property):
object.__setattr__(self, name, value)
if isinstance(value, framework.Parameter):
params = self.__dict__.get('_parameters', None)
if params is None:
......@@ -226,6 +228,11 @@ class Layer(core.Layer):
tensor = var.get_tensor()
tensor.set(self._loaddict_holder[value.name].numpy(),
framework._current_expected_place())
if name in params:
# remove unused param in tracer
if framework._dygraph_tracer_ is not None:
framework._dygraph_tracer_._vars.pop(params[name].name,
None)
params[name] = value
elif isinstance(value, core.Layer):
layers = self.__dict__.get('_sub_layers', None)
......
python/paddle/fluid/dygraph/nn.py
浏览文件 @ 54e07994
......@@ -83,7 +83,7 @@ class Conv2D(layers.Layer):
H_{out}&= \\frac{(H_{in} + 2 * paddings[0] - (dilations[0] * (H_f - 1) + 1))}{strides[0]} + 1 \\\\
W_{out}&= \\frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]} + 1
Args:
Parameters:
name_scope(str) : The name for this class.
num_filters(int): The number of filter. It is as same as the output
image channel.
......@@ -119,6 +119,10 @@ class Conv2D(layers.Layer):
act (str): Activation type, if it is set to None, activation is not appended.
Default: None
Attributes:
weight (Parameter): the learnable weights of filter of this layer.
bias (Parameter|None): the learnable bias of this layer.
Raises:
ValueError: If the shapes of input, filter_size, stride, padding and
groups mismatch.
......@@ -205,6 +209,22 @@ class Conv2D(layers.Layer):
dtype=self._dtype,
is_bias=True)
@property
def weight(self):
return self._filter_param
@weight.setter
def weight(self, value):
self._filter_param = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, input):
pre_bias = self._helper.create_variable_for_type_inference(
dtype=self._dtype)
......@@ -288,7 +308,7 @@ class Conv3D(layers.Layer):
H_{out}&= \\frac{(H_{in} + 2 * paddings[1] - (dilations[1] * (H_f - 1) + 1))}{strides[1]} + 1 \\\\
W_{out}&= \\frac{(W_{in} + 2 * paddings[2] - (dilations[2] * (W_f - 1) + 1))}{strides[2]} + 1
Args:
Parameters:
name_scope(str) : The name for this class.
num_filters(int): The number of filter. It is as same as the output image channel.
filter_size (int|tuple|None): The filter size. If filter_size is a tuple,
......@@ -323,6 +343,10 @@ class Conv3D(layers.Layer):
act (str): Activation type, if it is set to None, activation is not appended.
Default: None.
Attributes:
weight (Parameter): the learnable weights of filters of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: The tensor variable storing the convolution and \
non-linearity activation result.
......@@ -405,6 +429,22 @@ class Conv3D(layers.Layer):
dtype=self._dtype,
is_bias=True)
@property
def weight(self):
return self._filter_param
@weight.setter
def weight(self, value):
self._filter_param = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, input):
pre_bias = self._helper.create_variable_for_type_inference(
dtype=self._dtype)
......@@ -425,15 +465,17 @@ class Conv3D(layers.Layer):
'use_mkldnn': False
})
if self._bias_param is not None:
pre_act = self._helper.create_variable_for_type_inference(
dtype=self._dtype)
self._helper.append_op(
type='elementwise_add',
inputs={'X': [pre_bias],
'Y': [self._bias_param]},
outputs={'Out': [pre_act]},
attrs={'axis': 1})
else:
pre_act = pre_bias
return self._helper.append_activation(pre_act, act=self._act)
......@@ -489,7 +531,7 @@ class Conv3DTranspose(layers.Layer):
H_{out} &= (H_{in} - 1) * strides[1] - 2 * paddings[1] + dilations[1] * (H_f - 1) + 1 \\\\
W_{out} &= (W_{in} - 1) * strides[2] - 2 * paddings[2] + dilations[2] * (W_f - 1) + 1
Args:
Parameters:
name_scope(str) : The name for this class.
num_filters(int): The number of the filter. It is as same as the output
image channel.
......@@ -531,6 +573,10 @@ class Conv3DTranspose(layers.Layer):
name(str|None): A name for this layer(optional). If set None, the layer
will be named automatically.
Attributes:
weight (Parameter): the learnable weights of filters of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: The tensor variable storing the convolution transpose result.
......@@ -627,6 +673,22 @@ class Conv3DTranspose(layers.Layer):
dtype=self._dtype,
is_bias=True)
@property
def weight(self):
return self._img_filter
@weight.setter
def weight(self, value):
self._img_filter = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, input):
pre_bias = self._helper.create_variable_for_type_inference(
dtype=self._dtype)
......@@ -667,7 +729,7 @@ class Pool2D(layers.Layer):
Parameters(ksize, strides, paddings) are two elements. These two elements represent height and width, respectively.
The input(X) size and output(Out) size may be different.
Args:
Parameters:
name_scope(str) : The name of this class.
pool_size (int|list|tuple): The pool kernel size. If pool kernel size is a tuple or list,
it must contain two integers, (pool_size_Height, pool_size_Width).
......@@ -824,7 +886,7 @@ class FC(layers.Layer):
out.data = [[0.18669507, 0.1893476]]
out.shape = (1, 2)
Args:
Parameters:
name_scope(str): The name of this class.
size(int): The number of output units in this layer.
num_flatten_dims (int): The fc layer can accept an input tensor with more than
......@@ -845,6 +907,10 @@ class FC(layers.Layer):
is_test(bool): A flag indicating whether execution is in test phase. Default: False
dtype(str): Dtype used for weight
Attributes:
weight (list of Parameter): the learnable weights of this layer.
bias (Parameter|None): the learnable bias of this layer.
Raises:
ValueError: If rank of the input tensor is less than 2.
......@@ -883,15 +949,6 @@ class FC(layers.Layer):
self._act = act
self.__w = list()
@property
def _w(self, i=0):
return self.__w[i]
@_w.setter
def _w(self, value, i=0):
assert isinstance(value, Parameter)
self.__w[i] = value
def _build_once(self, input):
i = 0
for inp, param in self._helper.iter_inputs_and_params(input,
......@@ -916,6 +973,36 @@ class FC(layers.Layer):
self._b = self.create_parameter(
attr=self._bias_attr, shape=size, dtype=self._dtype, is_bias=True)
# TODO(songyouwei): We should remove _w property
@property
def _w(self, i=0):
return self.__w[i]
@_w.setter
def _w(self, value, i=0):
assert isinstance(self.__w[i], Variable)
self.__w[i].set_value(value)
@property
def weight(self):
if len(self.__w) > 1:
return self.__w
else:
return self.__w[0]
@weight.setter
def weight(self, value):
if len(self.__w) == 1:
self.__w[0] = value
@property
def bias(self):
return self._b
@bias.setter
def bias(self, value):
self._b = value
def forward(self, input):
mul_results = list()
i = 0
......@@ -1000,7 +1087,7 @@ class BatchNorm(layers.Layer):
\\sigma_{\\beta}^{2} + \\epsilon}} \\\\
y_i &\\gets \\gamma \\hat{x_i} + \\beta
Args:
Parameters:
name_scope(str): The name of this class.
act(str|None): Activation type, linear|relu|prelu|...
is_test (bool): A flag indicating whether it is in
......@@ -1184,7 +1271,7 @@ class Embedding(layers.Layer):
:attr:`input`.
All the input variables are passed in as local variables to the LayerHelper constructor
Args:
Parameters:
name_scope(str): The name of this class.
size(tuple|list): The shape of the look up table parameter. It should have two elements which indicate the size
of the dictionary of embeddings and the size of each embedding vector respectively.
......@@ -1196,6 +1283,9 @@ class Embedding(layers.Layer):
param_attr(ParamAttr): Parameters for this layer. Default: None.
dtype(np.dtype|core.VarDesc.VarType|str): The type of data : float32, float_16, int etc. Default: 'float32'.
Attributes:
weight (Parameter): the learnable weights of this layer.
Returns:
Variable: The tensor variable storing the embeddings of the \
supplied inputs.
......@@ -1246,6 +1336,14 @@ class Embedding(layers.Layer):
dtype=self._dtype,
is_bias=False)
@property
def weight(self):
return self._w
@weight.setter
def weight(self, value):
self._w = value
def forward(self, input):
out = self._helper.create_variable_for_type_inference(self._dtype)
self._helper.append_op(
......@@ -1291,7 +1389,7 @@ class LayerNorm(layers.Layer):
* :math:`b`: the trainable bias parameter.
Args:
Parameters:
name_scope(str): The name of this class.
scale(bool): Whether to learn the adaptive gain :math:`g` after
normalization. Default: True.
......@@ -1451,7 +1549,7 @@ class GRUUnit(layers.Layer):
This layer has three outputs :math:`h_t`, :math:`dot(r_t, h_{t-1})`
and concatenation of :math:`u_t`, :math:`r_t` and :math:`m_t`.
Args:
Parameters:
name_scope(str): The name of this class.
size (int): The input dimension value.
param_attr(ParamAttr|None): The parameter attribute for the learnable
......@@ -1481,6 +1579,10 @@ class GRUUnit(layers.Layer):
Default: 'sigmoid'
dtype(str): The dtype of the layers. Default: 'float32'
Attributes:
weight (Parameter): the learnable weights of this layer.
bias (Parameter): the learnable bias of this layer.
Returns:
tuple: The hidden value, reset-hidden value and gate values.
......@@ -1515,6 +1617,7 @@ class GRUUnit(layers.Layer):
origin_mode=False,
dtype='float32'):
super(GRUUnit, self).__init__(name_scope, dtype)
self._bias_attr = bias_attr
activation_dict = dict(
identity=0,
......@@ -1532,9 +1635,26 @@ class GRUUnit(layers.Layer):
# create bias
bias_size = [1, 3 * size]
self._bias_size = bias_size
self._bias = self.create_parameter(
attr=bias_attr, shape=bias_size, dtype=dtype, is_bias=True)
@property
def weight(self):
return self._weight
@weight.setter
def weight(self, value):
self._weight = value
@property
def bias(self):
return self._bias
@bias.setter
def bias(self, value):
self._bias = value
def forward(self, input, hidden):
inputs = {'Input': input, 'HiddenPrev': hidden, 'Weight': self._weight}
if self._bias:
......@@ -1567,7 +1687,7 @@ class NCE(layers.Layer):
`Noise-contrastive estimation: A new estimation principle for unnormalized statistical models <http://www.jmlr.org/proceedings/papers/v9/gutmann10a/gutmann10a.pdf>`_ .
By default this operator uses a uniform distribution for sampling.
Args:
Parameters:
name_scope(str): The name of this class.
num_total_classes (int): Total number of classes in all samples
param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights
......@@ -1590,6 +1710,10 @@ class NCE(layers.Layer):
seed (int): The seed used in sampler. Default: 0.
is_sparse(bool): The flag indicating whether to use sparse update, the weight@GRAD and bias@GRAD will be changed to SelectedRows. Default: False.
Attributes:
weight (Parameter): the learnable weights of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: The output nce loss.
......@@ -1763,6 +1887,22 @@ class NCE(layers.Layer):
self._inputs['Bias'] = self._b
self._inputs['Weight'] = self._w
@property
def weight(self):
return self._w
@weight.setter
def weight(self, value):
self._w = value
@property
def bias(self):
return self._b
@bias.setter
def bias(self, value):
self._b = value
def forward(self, input, label, sample_weight=None):
assert isinstance(input, Variable)
assert isinstance(label, Variable)
......@@ -1797,7 +1937,7 @@ class PRelu(layers.Layer):
.. math::
y = \max(0, x) + \\alpha * \min(0, x)
Args:
Parameters:
name_scope(str): The name of this class.
mode (str): The mode for weight sharing. It supports all, channel
and element. all: all elements share same weight
......@@ -1806,6 +1946,9 @@ class PRelu(layers.Layer):
param_attr(ParamAttr|None): The parameter attribute for the learnable
weight (alpha).
Attributes:
weight (Parameter): the learnable weights of this layer.
Returns:
Variable: The output tensor with the same shape as input.
......@@ -1849,6 +1992,14 @@ class PRelu(layers.Layer):
is_bias=False,
default_initializer=Constant(1.0))
@property
def weight(self):
return self._alpha
@weight.setter
def weight(self, value):
self._alpha = value
def forward(self, input):
out = self._helper.create_variable_for_type_inference(self._dtype)
......@@ -1878,7 +2029,7 @@ class BilinearTensorProduct(layers.Layer):
- :math:`out_{i}`: the i-th element of out, shape is [batch_size, size].
- :math:`y^\mathrm{T}`: the transpose of :math:`y_{2}`.
Args:
Parameters:
name_scope(str): The name of this class.
size (int): The dimension of this layer.
act (str): Activation to be applied to the output of this layer. Default: None.
......@@ -1889,6 +2040,10 @@ class BilinearTensorProduct(layers.Layer):
of this layer. If it is set to False, no bias will be added to the output units.
If it is set to None, the bias is initialized zero. Default: None.
Attributes:
weight (Parameter): the learnable weights of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: A 2-D Tensor of shape [batch_size, size].
......@@ -1940,6 +2095,22 @@ class BilinearTensorProduct(layers.Layer):
dtype=self._dtype,
is_bias=True)
@property
def weight(self):
return self._w
@weight.setter
def weight(self, value):
self._w = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, x, y):
self._inputs = {"X": x, "Y": y, "Weight": self._w}
if self._bias_param:
......@@ -2013,7 +2184,7 @@ class Conv2DTranspose(layers.Layer):
H_{out} &\in [ H^\prime_{out}, H^\prime_{out} + strides[0] ) \\\\
W_{out} &\in [ W^\prime_{out}, W^\prime_{out} + strides[1] )
Args:
Parameters:
name_scope(str): The name of this class.
num_filters(int): The number of the filter. It is as same as the output
image channel.
......@@ -2055,6 +2226,10 @@ class Conv2DTranspose(layers.Layer):
act (str): Activation type, if it is set to None, activation is not appended.
Default: None.
Attributes:
weight (Parameter): the learnable weights of filters of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: The tensor variable storing the convolution transpose result.
......@@ -2163,6 +2338,22 @@ class Conv2DTranspose(layers.Layer):
dtype=self._dtype,
is_bias=True)
@property
def weight(self):
return self._img_filter
@weight.setter
def weight(self, value):
self._img_filter = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, input):
pre_bias = self._helper.create_variable_for_type_inference(
dtype=input.dtype)
......@@ -2202,7 +2393,7 @@ class SequenceConv(layers.Layer):
other convolutional configurations for the filters and stride as given
in the input parameters to the function.
Args:
Parameters:
name_scope(str): The name of this class.
num_filters (int): number of filters.
filter_size (int): the filter size (H and W). Default: 3.
......@@ -2220,6 +2411,10 @@ class SequenceConv(layers.Layer):
act (str): Activation type, if it is set to None, activation is not appended.
Default: None.
Attributes:
weight (Parameter): the learnable weights of filters of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
Variable: output of sequence_conv
"""
......@@ -2305,7 +2500,7 @@ class RowConv(layers.Layer):
More details about row_conv please refer to the design document https://github.com/PaddlePaddle/Paddle/issues/2228#issuecomment-303903645 .
Args:
Parameters:
name_scope(str): The name of this class.
future_context_size (int): Future context size. Please note, the shape
of convolution kernel is [future_context_size + 1, D].
......@@ -2313,6 +2508,9 @@ class RowConv(layers.Layer):
name, initializer etc. Default: None.
act (str): Non-linear activation to be applied to output variable. Default: None.
Attributes:
weight (Parameter): the learnable weights of this layer.
Returns:
the output(Out) is a LodTensor, which supports variable time-length input sequences.
The underlying tensor in this LodTensor is a matrix with shape T x N, i.e., the same shape as X.
......@@ -2368,7 +2566,7 @@ class GroupNorm(layers.Layer):
Refer to `Group Normalization <https://arxiv.org/abs/1803.08494>`_ .
Args:
Parameters:
name_scope(str): The name of this class.
groups(int): The number of groups that divided from channels.
epsilon(float): The small value added to the variance to prevent
......@@ -2496,7 +2694,7 @@ class SpectralNorm(layers.Layer):
Refer to `Spectral Normalization <https://arxiv.org/abs/1802.05957>`_ .
Args:
Parameters:
name_scope(str): The name of this class.
dim(int): The index of dimension which should be permuted to the first before reshaping Input(Weight) to matrix, it should be set as 0 if Input(Weight) is the weight of fc layer, and should be set as 1 if Input(Weight) is the weight of conv layer. Default: 0.
power_iters(int): The number of power iterations to calculate spectral norm. Default: 1.
......@@ -2573,7 +2771,7 @@ class TreeConv(layers.Layer):
The paper of Tree-Based Convolution Operator is here: https://arxiv.org/abs/1409.5718v1
Args:
Parameters:
name_scope(str): The name of this class.
output_size(int): output feature width
num_filters(int): number of filters, Default: 1.
......@@ -2583,6 +2781,10 @@ class TreeConv(layers.Layer):
bias_attr(ParamAttr): the parameter attribute for the bias of this layer, Default: None.
name(str): a name of this layer(optional). If set None, the layer will be named automatically, Default: None.
Attributes:
weight (Parameter): the learnable weights of filters of this layer.
bias (Parameter|None): the learnable bias of this layer.
Returns:
out(Variable): (Tensor) The feature vector of subtrees. The shape of the output tensor is [max_tree_node_size, output_size, num_filters]. The output tensor could be a new feature vector for next tree convolution layers
......@@ -2639,6 +2841,22 @@ class TreeConv(layers.Layer):
dtype=self._dtype,
is_bias=False)
@property
def weight(self):
return self.W
@weight.setter
def weight(self, value):
self.W = value
@property
def bias(self):
return self._bias_param
@bias.setter
def bias(self, value):
self._bias_param = value
def forward(self, nodes_vector, edge_set):
if self._name:
......
python/paddle/fluid/framework.py
浏览文件 @ 54e07994
......@@ -638,6 +638,45 @@ class Variable(object):
new_ivar = self._ivar._copy_to(core.CPUPlace(), True)
return np.array(new_ivar.value().get_tensor())
@dygraph_only
def set_value(self, value):
"""
Set a new value for this Variable.
Args:
value (Variable|np.ndarray): the new value.
Returns:
None.
Examples:
.. code-block:: python
import paddle.fluid as fluid
from paddle.fluid.dygraph.base import to_variable
from paddle.fluid.dygraph import FC
import numpy as np
data = np.ones([3, 32, 32], dtype='float32')
with fluid.dygraph.guard():
fc = fluid.dygraph.FC("fc", 4)
t = to_variable(data)
fc(t) # call with default weight
custom_weight = np.random.randn(1024, 4).astype("float32")
fc.weight.set_value(custom_weight) # change existing weight
out = fc(t) # call with different weight
"""
assert isinstance(value, (Variable, np.ndarray))
if list(value.shape) != list(self.shape):
raise ValueError(
"The shape of the new value must be the same as that of the original Variable."
)
self_tensor = self._ivar.value().get_tensor()
if isinstance(value, Variable):
value = value._ivar.value().get_tensor().__array__()
self_tensor.set(value, _current_expected_place())
@dygraph_only
def backward(self, backward_strategy=None):
"""
......@@ -1042,7 +1081,7 @@ class Variable(object):
if self.shape[axis] < 0:
return self._cloneVar(True)
index = int(item)
if (index > 0 and index >= self.shape[axis])\
if (index > 0 and index >= self.shape[axis]) \
or (index < 0 and (index + self.shape[axis]) < 0):
raise IndexError("invalid index")
return self._sliceVar([axis], [index], [index + 1])
......
python/paddle/fluid/tests/unittests/test_layers.py
浏览文件 @ 54e07994
......@@ -82,6 +82,34 @@ class LayerTest(unittest.TestCase):
class TestLayer(LayerTest):
def test_custom_layer_with_kwargs(self):
class CustomLayer(fluid.Layer):
def __init__(self, name_scope, fc1_size=4):
super(CustomLayer, self).__init__(name_scope)
self.fc1 = nn.FC('fc1',
size=fc1_size,
bias_attr=False,
num_flatten_dims=1)
self.fc2 = nn.FC('fc2',
size=1,
bias_attr=False,
num_flatten_dims=1)
def forward(self, x, do_fc2=False):
ret = self.fc1(x)
if do_fc2:
ret = self.fc2(ret)
return ret
with self.dynamic_graph():
inp = np.ones([3, 3], dtype='float32')
x = base.to_variable(inp)
custom = CustomLayer('custom', fc1_size=2)
ret = custom(x, do_fc2=False)
self.assertTrue(np.array_equal(ret.numpy().shape, [3, 2]))
ret = custom(x, do_fc2=True)
self.assertTrue(np.array_equal(ret.numpy().shape, [3, 1]))
def test_fc(self):
inp = np.ones([3, 32, 32], dtype='float32')
with self.static_graph():
......@@ -117,6 +145,62 @@ class TestLayer(LayerTest):
self.assertTrue(np.array_equal(static_ret, static_ret2))
self.assertTrue(np.array_equal(static_ret, dy_ret_value))
with self.dynamic_graph():
custom_weight = np.random.randn(1024, 4).astype("float32")
weight_attr1 = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
fc1 = fluid.dygraph.FC("fc1",
4,
num_flatten_dims=1,
param_attr=weight_attr1)
out1 = fc1(base.to_variable(inp))
loss1 = fluid.layers.reduce_mean(out1)
fc1_weight_init = fc1.weight.detach()
fc1_bias_init = fc1.bias.detach()
loss1.backward()
optimizer1 = fluid.optimizer.SGD(learning_rate=0.1)
optimizer1.minimize(loss1)
fc1_weight_updated = fc1.weight.detach()
with self.dynamic_graph():
weight_attr2 = fluid.ParamAttr(
initializer=fluid.initializer.Uniform())
fc2 = fluid.dygraph.FC("fc2",
4,
num_flatten_dims=1,
param_attr=weight_attr2)
out2 = fc2(base.to_variable(inp))
self.assertFalse(
np.array_equal(fc1_weight_init.numpy(), fc2.weight.numpy()))
self.assertFalse(np.array_equal(out1.numpy(), out2.numpy()))
mismatched_weight = np.random.randn(4, 4).astype("float32")
with self.assertRaises(ValueError):
fc2.weight.set_value(mismatched_weight)
fc2.weight.set_value(fc1_weight_init)
fc2.bias.set_value(fc1_bias_init)
out2 = fc2(base.to_variable(inp))
loss2 = fluid.layers.reduce_mean(out2)
loss2.backward()
optimizer2 = fluid.optimizer.SGD(learning_rate=0.1)
optimizer2.minimize(loss2)
self.assertTrue(
np.array_equal(fc2.weight.numpy(), fc1_weight_updated.numpy()))
self.assertTrue(np.array_equal(out1.numpy(), out2.numpy()))
fc2.weight = fc1.weight
fc2.bias = fc1.bias
self.assertTrue(
np.array_equal(fc2.weight.numpy(), fc1.weight.numpy()))
self.assertTrue(np.array_equal(fc2.bias.numpy(), fc1.bias.numpy()))
def test_layer_norm(self):
inp = np.ones([3, 32, 32], dtype='float32')
with self.static_graph():
......@@ -238,6 +322,41 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_ret, dy_ret_value))
self.assertTrue(np.allclose(static_ret, static_ret2))
with self.dynamic_graph():
images = np.ones([2, 3, 5, 5], dtype='float32')
custom_weight = np.random.randn(3, 3, 2, 2).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
conv2d1 = nn.Conv2D('conv2d1', num_filters=3, filter_size=[2, 2])
conv2d2 = nn.Conv2D(
'conv2d2',
num_filters=3,
filter_size=[2, 2],
param_attr=weight_attr)
dy_ret1 = conv2d1(base.to_variable(images))
dy_ret2 = conv2d2(base.to_variable(images))
self.assertFalse(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv2d1_weight_np = conv2d1.weight.numpy()
conv2d1_bias = conv2d1.bias
self.assertFalse(
np.array_equal(conv2d1_weight_np, conv2d2.weight.numpy()))
conv2d2.weight.set_value(conv2d1_weight_np)
self.assertTrue(
np.array_equal(conv2d1_weight_np, conv2d2.weight.numpy()))
conv2d2.bias.set_value(conv2d1_bias)
dy_ret1 = conv2d1(base.to_variable(images))
dy_ret2 = conv2d2(base.to_variable(images))
self.assertTrue(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv2d2.weight = conv2d1.weight
conv2d2.bias = conv2d1.bias
self.assertTrue(
np.array_equal(conv2d1.weight.numpy(), conv2d2.weight.numpy()))
self.assertTrue(
np.array_equal(conv2d1.bias.numpy(), conv2d2.bias.numpy()))
def test_gru_unit(self):
lod = [[2, 4, 3]]
D = 5
......@@ -282,6 +401,37 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_ret[i], static_ret2[i]))
self.assertTrue(np.allclose(static_ret[i], dy_ret_value[i]))
with self.dynamic_graph():
custom_weight = np.random.randn(D, D * 3).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
gru1 = nn.GRUUnit('gru1', size=D * 3)
gru2 = nn.GRUUnit('gru2', size=D * 3, param_attr=weight_attr)
dy_ret1 = gru1(
base.to_variable(input), base.to_variable(hidden_input))
dy_ret2 = gru2(
base.to_variable(input), base.to_variable(hidden_input))
self.assertFalse(
np.array_equal(gru1.weight.numpy(), gru2.weight.numpy()))
for o1, o2 in zip(dy_ret1, dy_ret2):
self.assertFalse(np.array_equal(o1.numpy(), o2.numpy()))
gru2.weight.set_value(gru1.weight.numpy())
gru2.bias.set_value(gru1.bias)
dy_ret1 = gru1(
base.to_variable(input), base.to_variable(hidden_input))
dy_ret2 = gru2(
base.to_variable(input), base.to_variable(hidden_input))
for o1, o2 in zip(dy_ret1, dy_ret2):
self.assertTrue(np.array_equal(o1.numpy(), o2.numpy()))
gru2.weight = gru1.weight
gru2.bias = gru1.bias
self.assertTrue(
np.array_equal(gru1.weight.numpy(), gru2.weight.numpy()))
self.assertTrue(
np.array_equal(gru1.bias.numpy(), gru2.bias.numpy()))
def test_elementwise_math(self):
n = np.ones([3, 3], dtype='float32')
n2 = np.ones([3, 3], dtype='float32') * 1.1
......@@ -417,6 +567,42 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_rlt2, static_rlt))
self.assertTrue(np.allclose(dy_rlt_value, static_rlt2))
with self.dynamic_graph():
images = np.ones([2, 3, 5, 5], dtype='float32')
custom_weight = np.random.randn(3, 3, 2, 2).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
conv2d1 = nn.Conv2DTranspose(
'conv2d1', num_filters=3, filter_size=[2, 2])
conv2d2 = nn.Conv2DTranspose(
'conv2d2',
num_filters=3,
filter_size=[2, 2],
param_attr=weight_attr)
dy_ret1 = conv2d1(base.to_variable(images))
dy_ret2 = conv2d2(base.to_variable(images))
self.assertFalse(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv2d1_weight_np = conv2d1.weight.numpy()
conv2d1_bias = conv2d1.bias
self.assertFalse(
np.array_equal(conv2d1_weight_np, conv2d2.weight.numpy()))
conv2d2.weight.set_value(conv2d1_weight_np)
self.assertTrue(
np.array_equal(conv2d1_weight_np, conv2d2.weight.numpy()))
conv2d2.bias.set_value(conv2d1_bias)
dy_ret1 = conv2d1(base.to_variable(images))
dy_ret2 = conv2d2(base.to_variable(images))
self.assertTrue(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv2d2.weight = conv2d1.weight
conv2d2.bias = conv2d1.bias
self.assertTrue(
np.array_equal(conv2d1.weight.numpy(), conv2d2.weight.numpy()))
self.assertTrue(
np.array_equal(conv2d1.bias.numpy(), conv2d2.bias.numpy()))
def test_bilinear_tensor_product(self):
inp_np_x = np.array([[1, 2, 3]]).astype('float32')
inp_np_y = np.array([[4, 5, 6]]).astype('float32')
......@@ -498,9 +684,36 @@ class TestLayer(LayerTest):
self.assertTrue(np.array_equal(static_rlt2, static_rlt))
self.assertTrue(np.array_equal(dy_rlt_value, static_rlt))
with self.dynamic_graph():
custom_weight = np.random.randn(6, 3, 3).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
btp1 = nn.BilinearTensorProduct('btp1', 6, act='sigmoid')
btp2 = nn.BilinearTensorProduct(
'btp2', 6, act='sigmoid', param_attr=weight_attr)
dy_rlt1 = btp1(
base.to_variable(inp_np_x), base.to_variable(inp_np_y))
dy_rlt2 = btp2(
base.to_variable(inp_np_x), base.to_variable(inp_np_y))
self.assertFalse(np.array_equal(dy_rlt1.numpy(), dy_rlt2.numpy()))
btp2.weight.set_value(btp1.weight.numpy())
btp2.bias.set_value(btp1.bias)
dy_rlt1 = btp1(
base.to_variable(inp_np_x), base.to_variable(inp_np_y))
dy_rlt2 = btp2(
base.to_variable(inp_np_x), base.to_variable(inp_np_y))
self.assertTrue(np.array_equal(dy_rlt1.numpy(), dy_rlt2.numpy()))
btp2.weight = btp1.weight
btp2.bias = btp1.bias
self.assertTrue(
np.array_equal(btp1.weight.numpy(), btp2.weight.numpy()))
self.assertTrue(
np.array_equal(btp1.bias.numpy(), btp2.bias.numpy()))
def test_prelu(self):
inp_np = np.ones([5, 200, 100, 100]).astype('float32')
with self.static_graph():
data_t = layers.data(
name="input",
......@@ -540,6 +753,32 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_rlt2, static_rlt))
self.assertTrue(np.allclose(dy_rlt_value, static_rlt))
with self.dynamic_graph():
inp_np = np.random.randn(5, 200, 100, 100).astype("float32")
inp = base.to_variable(inp_np)
mode = 'channel'
prelu1 = nn.PRelu(
'prelu1',
mode=mode,
param_attr=ParamAttr(initializer=Constant(2.0)))
prelu2 = nn.PRelu(
'prelu2',
mode=mode,
param_attr=ParamAttr(initializer=Constant(1.0)))
dy_rlt1 = prelu1(inp)
dy_rlt2 = prelu2(inp)
self.assertFalse(
np.array_equal(prelu1.weight.numpy(), prelu2.weight.numpy()))
self.assertFalse(np.array_equal(dy_rlt1.numpy(), dy_rlt2.numpy()))
prelu2.weight.set_value(prelu1.weight.numpy())
dy_rlt1 = prelu1(inp)
dy_rlt2 = prelu2(inp)
self.assertTrue(np.array_equal(dy_rlt1.numpy(), dy_rlt2.numpy()))
prelu2.weight = prelu1.weight
self.assertTrue(
np.array_equal(prelu1.weight.numpy(), prelu2.weight.numpy()))
def test_embeding(self):
inp_word = np.array([[[1]]]).astype('int64')
dict_size = 20
......@@ -574,6 +813,31 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_rlt2, static_rlt))
self.assertTrue(np.allclose(dy_rlt_value, static_rlt))
with self.dynamic_graph():
custom_weight = np.random.randn(dict_size, 32).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
emb1 = nn.Embedding(
name_scope='embedding', size=[dict_size, 32], is_sparse=False)
emb2 = nn.Embedding(
name_scope='embedding',
size=[dict_size, 32],
param_attr=weight_attr,
is_sparse=False)
rep1 = emb1(base.to_variable(inp_word))
rep2 = emb2(base.to_variable(inp_word))
self.assertFalse(np.array_equal(emb1.weight.numpy(), custom_weight))
self.assertTrue(np.array_equal(emb2.weight.numpy(), custom_weight))
self.assertFalse(np.array_equal(rep1.numpy(), rep2.numpy()))
emb2.weight.set_value(emb1.weight.numpy())
rep2 = emb2(base.to_variable(inp_word))
self.assertTrue(np.array_equal(rep1.numpy(), rep2.numpy()))
emb2.weight = emb1.weight
self.assertTrue(
np.array_equal(emb1.weight.numpy(), emb2.weight.numpy()))
def test_nce(self):
window_size = 5
dict_size = 20
......@@ -695,6 +959,69 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_rlt2, static_rlt))
self.assertTrue(np.allclose(dy_rlt_value, static_rlt))
with self.dynamic_graph(force_to_use_cpu=True):
custom_weight = np.random.randn(dict_size, 128).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
words = []
for i in range(window_size):
words.append(base.to_variable(inp_word[i]))
sample_weights = layers.fill_constant(
shape=[5, 1], dtype='float32', value=1)
emb = nn.Embedding(
'embedding',
size=[dict_size, 32],
param_attr='emb.w',
is_sparse=False)
embs3 = []
for i in range(window_size):
if i == label_word:
continue
emb_rlt = emb(words[i])
embs3.append(emb_rlt)
embs3 = layers.concat(input=embs3, axis=1)
nce1 = nn.NCE('nce1',
num_total_classes=dict_size,
num_neg_samples=2,
sampler="custom_dist",
custom_dist=nid_freq_arr.tolist(),
seed=seed,
param_attr='nce1.w',
bias_attr='nce1.b',
sample_weight=sample_weights)
nce2 = nn.NCE('nce2',
param_attr=weight_attr,
num_total_classes=dict_size,
num_neg_samples=2,
sampler="custom_dist",
custom_dist=nid_freq_arr.tolist(),
seed=seed,
bias_attr='nce2.b',
sample_weight=sample_weights)
nce1_loss = nce1(embs3, words[label_word])
nce2_loss = nce2(embs3, words[label_word])
self.assertFalse(
np.array_equal(nce1_loss.numpy(), nce2_loss.numpy()))
nce2.weight.set_value(nce1.weight.numpy())
nce2.bias.set_value(nce1.bias)
nce1_loss = nce1(embs3, words[label_word])
nce2_loss = nce2(embs3, words[label_word])
self.assertTrue(
np.array_equal(nce1_loss.numpy(), nce2_loss.numpy()))
nce2.weight = nce1.weight
nce2.bias = nce1.bias
self.assertTrue(
np.array_equal(nce1.weight.numpy(), nce2.weight.numpy()))
self.assertTrue(
np.array_equal(nce1.bias.numpy(), nce2.bias.numpy()))
def test_conv3d(self):
with self.static_graph():
images = layers.data(
......@@ -724,6 +1051,38 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_ret, dy_rlt_value))
self.assertTrue(np.allclose(static_ret, static_ret2))
with self.dynamic_graph():
images = np.ones([2, 3, 6, 6, 6], dtype='float32')
custom_weight = np.random.randn(3, 3, 2, 2, 2).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
conv3d1 = nn.Conv3D('conv3d1', num_filters=3, filter_size=2)
conv3d2 = nn.Conv3D(
'conv3d2', num_filters=3, filter_size=2, param_attr=weight_attr)
dy_ret1 = conv3d1(base.to_variable(images))
dy_ret2 = conv3d2(base.to_variable(images))
self.assertFalse(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv3d1_weight_np = conv3d1.weight.numpy()
conv3d1_bias = conv3d1.bias
self.assertFalse(
np.array_equal(conv3d1_weight_np, conv3d2.weight.numpy()))
conv3d2.weight.set_value(conv3d1_weight_np)
self.assertTrue(
np.array_equal(conv3d1_weight_np, conv3d2.weight.numpy()))
conv3d1.bias.set_value(conv3d1_bias)
dy_ret1 = conv3d1(base.to_variable(images))
dy_ret2 = conv3d2(base.to_variable(images))
self.assertTrue(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv3d2.weight = conv3d1.weight
conv3d2.bias = conv3d1.bias
self.assertTrue(
np.array_equal(conv3d1.weight.numpy(), conv3d2.weight.numpy()))
self.assertTrue(
np.array_equal(conv3d1.bias.numpy(), conv3d2.bias.numpy()))
def test_row_conv(self):
input = np.arange(15).reshape([3, 5]).astype('float32')
if core.is_compiled_with_cuda():
......@@ -943,6 +1302,45 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_ret, static_ret2))
self.assertTrue(np.allclose(static_ret, dy_rlt_value))
with self.dynamic_graph():
custom_weight = np.random.randn(5, 3, 6, 1).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
treeConv1 = nn.TreeConv(
'SpectralNorm1',
output_size=6,
num_filters=1,
max_depth=2,
bias_attr='tc1_b')
treeConv2 = nn.TreeConv(
'SpectralNorm2',
output_size=6,
num_filters=1,
max_depth=2,
param_attr=weight_attr,
bias_attr='tc2_b')
dy_ret1 = treeConv1(
base.to_variable(vectors), base.to_variable(adj))
dy_ret2 = treeConv2(
base.to_variable(vectors), base.to_variable(adj))
self.assertFalse(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
treeConv2.weight.set_value(treeConv1.weight.numpy())
treeConv2.bias.set_value(treeConv1.bias)
dy_ret1 = treeConv1(
base.to_variable(vectors), base.to_variable(adj))
dy_ret2 = treeConv2(
base.to_variable(vectors), base.to_variable(adj))
self.assertTrue(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
treeConv2.weight = treeConv1.weight
treeConv2.bias = treeConv1.bias
self.assertTrue(
np.array_equal(treeConv1.weight.numpy(),
treeConv2.weight.numpy()))
self.assertTrue(
np.array_equal(treeConv1.bias.numpy(), treeConv2.bias.numpy()))
def test_conv3d_transpose(self):
input_array = np.arange(0, 48).reshape(
[2, 3, 2, 2, 2]).astype('float32')
......@@ -974,6 +1372,48 @@ class TestLayer(LayerTest):
self.assertTrue(np.allclose(static_rlt2, static_rlt))
self.assertTrue(np.allclose(dy_rlt_value, static_rlt))
with self.dynamic_graph():
images = np.ones([2, 3, 6, 6, 6], dtype='float32')
custom_weight = np.random.randn(3, 3, 2, 2, 2).astype("float32")
weight_attr = fluid.ParamAttr(
initializer=fluid.initializer.NumpyArrayInitializer(
custom_weight))
conv3d1 = nn.Conv3DTranspose(
'conv3d1',
num_filters=3,
filter_size=2,
bias_attr='conv3d1_b',
use_cudnn=False)
conv3d2 = nn.Conv3DTranspose(
'conv3d2',
num_filters=3,
filter_size=2,
param_attr=weight_attr,
bias_attr='conv3d2_b',
use_cudnn=False)
dy_ret1 = conv3d1(base.to_variable(images))
dy_ret2 = conv3d2(base.to_variable(images))
self.assertFalse(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv3d1_weight_np = conv3d1.weight.numpy()
conv3d1_bias = conv3d1.bias
self.assertFalse(
np.array_equal(conv3d1_weight_np, conv3d2.weight.numpy()))
conv3d2.weight.set_value(conv3d1_weight_np)
self.assertTrue(
np.array_equal(conv3d1_weight_np, conv3d2.weight.numpy()))
conv3d1.bias.set_value(conv3d1_bias)
dy_ret1 = conv3d1(base.to_variable(images))
dy_ret2 = conv3d2(base.to_variable(images))
self.assertTrue(np.array_equal(dy_ret1.numpy(), dy_ret2.numpy()))
conv3d2.weight = conv3d1.weight
conv3d2.bias = conv3d1.bias
self.assertTrue(
np.array_equal(conv3d1.weight.numpy(), conv3d2.weight.numpy()))
self.assertTrue(
np.array_equal(conv3d1.bias.numpy(), conv3d2.bias.numpy()))
def test_eye_op(self):
np_eye = np.eye(3, 2)
array_rlt1 = [np_eye for _ in range(3)]
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或