Add version_info in model.config.

CHANGELOG.md

@@ -85,6 +85,16 @@ To release a new version, please update the changelog as followed:
 
 ### Contributors
 
+## [2.0.3]
+
+### Changed
+- Add version_info in model.config. 
+
+### Fixed
+
+### Contributors
+- @warshallrho:
+
 ## [2.0.2] - 2019-6-5
 
 ### Changed

tensorlayer/files/utils.py

@@ -181,21 +181,23 @@ def save_hdf5_graph(network, filepath='model.hdf5', save_weights=False, customiz
     logging.info("[*] Saving TL model into {}, saving weights={}".format(filepath, save_weights))
 
     model_config = network.config  # net2static_graph(network)
+    model_config["version_info"]["save_date"] = datetime.datetime.utcnow().replace(tzinfo=datetime.timezone.utc
+                                                                                  ).isoformat()
     model_config_str = str(model_config)
     customized_data_str = str(customized_data)
-    version_info = {
-        "tensorlayer_version": tl.__version__,
-        "backend": "tensorflow",
-        "backend_version": tf.__version__,
-        "training_device": "gpu",
-        "save_date": datetime.datetime.utcnow().replace(tzinfo=datetime.timezone.utc).isoformat()
-    }
-    version_info_str = str(version_info)
+    # version_info = {
+    #     "tensorlayer_version": tl.__version__,
+    #     "backend": "tensorflow",
+    #     "backend_version": tf.__version__,
+    #     "training_device": "gpu",
+    #     "save_date": datetime.datetime.utcnow().replace(tzinfo=datetime.timezone.utc).isoformat()
+    # }
+    # version_info_str = str(version_info)
 
     with h5py.File(filepath, 'w') as f:
         f.attrs["model_config"] = model_config_str.encode('utf8')
         f.attrs["customized_data"] = customized_data_str.encode('utf8')
-        f.attrs["version_info"] = version_info_str.encode('utf8')
+        # f.attrs["version_info"] = version_info_str.encode('utf8')
         if save_weights:
             _save_weights_to_hdf5_group(f, network.all_layers)
         f.flush()
@@ -315,8 +317,12 @@ def load_hdf5_graph(filepath='model.hdf5', load_weights=False):
 
     f = h5py.File(filepath, 'r')
 
-    version_info_str = f.attrs["version_info"].decode('utf8')
-    version_info = eval(version_info_str)
+    model_config_str = f.attrs["model_config"].decode('utf8')
+    model_config = eval(model_config_str)
+
+    # version_info_str = f.attrs["version_info"].decode('utf8')
+    # version_info = eval(version_info_str)
+    version_info = model_config["version_info"]
     backend_version = version_info["backend_version"]
     tensorlayer_version = version_info["tensorlayer_version"]
     if backend_version != tf.__version__:
@@ -332,9 +338,6 @@ def load_hdf5_graph(filepath='model.hdf5', load_weights=False):
             )
         )
 
-    model_config_str = f.attrs["model_config"].decode('utf8')
-    model_config = eval(model_config_str)
-
     M = static_graph2net(model_config)
     if load_weights:
         if not ('layer_names' in f.attrs.keys()):

tensorlayer/layers/recurrent.py

@@ -42,24 +42,31 @@ class RNN(Layer):
         A RNN cell implemented by tf.keras
             - E.g. tf.keras.layers.SimpleRNNCell, tf.keras.layers.LSTMCell, tf.keras.layers.GRUCell
             - Note TF2.0+, TF1.0+ and TF1.0- are different
+
     return_last_output : boolean
         Whether return last output or all outputs in a sequence.
+
             - If True, return the last output, "Sequence input and single output"
             - If False, return all outputs, "Synced sequence input and output"
             - In other word, if you want to stack more RNNs on this layer, set to False
+
         In a dynamic model, `return_last_output` can be updated when it is called in customised forward().
         By default, `False`.
     return_seq_2d : boolean
         Only consider this argument when `return_last_output` is `False`
+
             - If True, return 2D Tensor [batch_size * n_steps, n_hidden], for stacking Dense layer after it.
             - If False, return 3D Tensor [batch_size, n_steps, n_hidden], for stacking multiple RNN after it.
+
         In a dynamic model, `return_seq_2d` can be updated when it is called in customised forward().
         By default, `False`.
     return_last_state: boolean
         Whether to return the last state of the RNN cell. The state is a list of Tensor.
         For simple RNN and GRU, last_state = [last_output]; For LSTM, last_state = [last_output, last_cell_state]
+
             - If True, the layer will return outputs and the final state of the cell.
             - If False, the layer will return outputs only.
+
         In a dynamic model, `return_last_state` can be updated when it is called in customised forward().
         By default, `False`.
     in_channels: int
@@ -74,6 +81,7 @@ class RNN(Layer):
     For synced sequence input and output, see `PTB example <https://github.com/tensorlayer/tensorlayer/blob/master/example/tutorial_ptb_lstm.py>`__
 
     A simple regression model below.
+
     >>> inputs = tl.layers.Input([batch_size, num_steps, embedding_size])
     >>> rnn_out, lstm_state = tl.layers.RNN(
     >>>     cell=tf.keras.layers.LSTMCell(units=hidden_size, dropout=0.1),
@@ -85,6 +93,7 @@ class RNN(Layer):
     >>> # If LSTMCell is applied, the rnn_state is [h, c] where h the hidden state and c the cell state of LSTM.
 
     A stacked RNN model.
+
     >>> inputs = tl.layers.Input([batch_size, num_steps, embedding_size])
     >>> rnn_out1 = tl.layers.RNN(
     >>>     cell=tf.keras.layers.SimpleRNNCell(units=hidden_size, dropout=0.1),
@@ -101,7 +110,6 @@ class RNN(Layer):
     -----
     Input dimension should be rank 3 : [batch_size, n_steps, n_features], if no, please see layer :class:`Reshape`.
 
-
     """
 
     def __init__(
@@ -237,6 +245,7 @@ class BiRNN(Layer):
         Whether to return the last state of the two cells. The state is a list of Tensor.
             - If True, the layer will return outputs, the final state of `fw_cell` and the final state of `bw_cell`.
             - If False, the layer will return outputs only.
+
         In a dynamic model, `return_last_state` can be updated when it is called in customised forward().
         By default, `False`.
     in_channels: int
@@ -249,6 +258,7 @@ class BiRNN(Layer):
     Examples
     --------
     A simple regression model below.
+
     >>> inputs = tl.layers.Input([batch_size, num_steps, embedding_size])
     >>> # the fw_cell and bw_cell can be different
     >>> rnnlayer = tl.layers.BiRNN(
@@ -266,6 +276,7 @@ class BiRNN(Layer):
     >>> rnn_model = tl.models.Model(inputs=inputs, outputs=[outputs, rnn_out, rnn_fw_state[0], rnn_bw_state[0]])
 
     A stacked BiRNN model.
+
     >>> inputs = tl.layers.Input([batch_size, num_steps, embedding_size])
     >>> rnn_out1 = tl.layers.BiRNN(
     >>>     fw_cell=tf.keras.layers.SimpleRNNCell(units=hidden_size, dropout=0.1),
@@ -281,7 +292,6 @@ class BiRNN(Layer):
     >>> outputs = tl.layers.Reshape([-1, num_steps])(dense)
     >>> rnn_model = tl.models.Model(inputs=inputs, outputs=outputs)
 
-
     Notes
     -----
     Input dimension should be rank 3 : [batch_size, n_steps, n_features]. If not, please see layer :class:`Reshape`.
@@ -612,10 +622,12 @@ class ConvLSTM(Layer):
             - If True, return the last output, "Sequence input and single output".
             - If False, return all outputs, "Synced sequence input and output".
             - In other word, if you want to stack more RNNs on this layer, set to False.
+
     return_seq_2d : boolean
         Only consider this argument when `return_last_output` is `False`
             - If True, return 2D Tensor [n_example, n_hidden], for stacking DenseLayer after it.
             - If False, return 3D Tensor [n_example/n_steps, n_steps, n_hidden], for stacking multiple RNN after it.
+
     name : str
         A unique layer name.
 
@@ -886,6 +898,7 @@ class Seq2Seq(Layer):
         A TensorFlow core RNN cell
             - see `RNN Cells in TensorFlow <https://www.tensorflow.org/api_docs/python/>`__
             - Note TF1.0+ and TF1.0- are different
+
     cell_init_args : dictionary or None
         The arguments for the cell initializer.
     n_hidden : int
@@ -903,12 +916,14 @@ class Seq2Seq(Layer):
     dropout : tuple of float or int
         The input and output keep probability (input_keep_prob, output_keep_prob).
             - If one int, input and output keep probability are the same.
+
     n_layer : int
         The number of RNN layers, default is 1.
     return_seq_2d : boolean
         Only consider this argument when `return_last_output` is `False`
             - If True, return 2D Tensor [n_example, 2 * n_hidden], for stacking DenseLayer after it.
             - If False, return 3D Tensor [n_example/n_steps, n_steps, 2 * n_hidden], for stacking multiple RNN after it.
+
     name : str
         A unique layer name.
 

tensorlayer/models/core.py

@@ -443,13 +443,14 @@ def config(self):
                 _config.update({"name": None})
             else:
                 _config.update({"name": self.name})
-            # versionInfo = {
-            #     "tensorlayer_version": tl.__version__,
-            #     "backend": "tensorflow",
-            #     "backend_version": tf.__version__,
-            #     "training_device": "gpu",
-            # }
-            # _config.update(versionInfo)
+            version_info = {
+                "tensorlayer_version": tl.__version__,
+                "backend": "tensorflow",
+                "backend_version": tf.__version__,
+                "training_device": "gpu",
+                "save_date": None,
+            }
+            _config["version_info"] = version_info
             # if self.outputs is None:
             #     raise RuntimeError(
             #         "Dynamic mode does not support config yet."

tests/models/test_model_save_graph.py

@@ -15,6 +15,11 @@
 from tests.utils import CustomTestCase
 
 
+def RemoveDateInConfig(config):
+    config["version_info"]["save_date"] = None
+    return config
+
+
 def basic_static_model():
     ni = Input((None, 24, 24, 3))
     nn = Conv2d(16, (5, 5), (1, 1), padding='SAME', act=tf.nn.relu, name="conv1")(ni)
@@ -42,7 +47,10 @@ def test_save(self):
         M1.save(filepath="basic_model_without_weights.hdf5", save_weights=False)
         M2 = Model.load(filepath="basic_model_without_weights.hdf5", load_weights=False)
 
-        self.assertEqual(M1.config, M2.config)
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
+        self.assertEqual(M1_config, M2_config)
 
 
 def get_model(inputs_shape):
@@ -188,7 +196,10 @@ def test_save(self):
         M1.save(filepath="siamese.hdf5", save_weights=False)
         M2 = Model.load(filepath="siamese.hdf5", load_weights=False)
 
-        self.assertEqual(M1.config, M2.config)
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
+        self.assertEqual(M1_config, M2_config)
 
 
 class Vgg_LayerList_test(CustomTestCase):
@@ -204,7 +215,10 @@ def test_save(self):
         M1.save(filepath="vgg.hdf5", save_weights=False)
         M2 = Model.load(filepath="vgg.hdf5", load_weights=False)
 
-        self.assertEqual(M1.config, M2.config)
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
+        self.assertEqual(M1_config, M2_config)
 
 
 class List_inputs_outputs_test(CustomTestCase):
@@ -228,7 +242,10 @@ def test_list_inputs_outputs(self):
         M1.save(filepath="list.hdf5", save_weights=False)
         M2 = Model.load(filepath="list.hdf5", load_weights=False)
 
-        self.assertEqual(M1.config, M2.config)
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
+        self.assertEqual(M1_config, M2_config)
 
 
 class Lambda_layer_test(CustomTestCase):
@@ -251,8 +268,11 @@ def test_lambda_layer_no_para_no_args(self):
         output1 = M1(npInput).numpy()
         output2 = M1(npInput).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
         self.assertEqual((output1 == output2).all(), True)
-        self.assertEqual(M1.config, M2.config)
+        self.assertEqual(M1_config, M2_config)
 
     def test_lambda_layer_no_para_with_args(self):
 
@@ -272,9 +292,12 @@ def customize_func(x, foo=42):  # x is the inputs, foo is an argument
         output1 = M1(npInput).numpy()
         output2 = M2(npInput).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
         self.assertEqual((output1 == output2).all(), True)
         self.assertEqual((output1 == (np.zeros((8, 3)) + 9)).all(), True)
-        self.assertEqual(M1.config, M2.config)
+        self.assertEqual(M1_config, M2_config)
 
     def test_lambda_layer_keras_model(self):
         input_shape = [100, 5]
@@ -299,8 +322,11 @@ def test_lambda_layer_keras_model(self):
         output2 = M2(npInput).numpy()
         output4 = M4(npInput).numpy()
 
+        M2_config = RemoveDateInConfig(M2.config)
+        M4_config = RemoveDateInConfig(M4.config)
+
         self.assertEqual((output2 == output4).all(), True)
-        self.assertEqual(M2.config, M4.config)
+        self.assertEqual(M2_config, M4_config)
 
         ori_weights = M4.all_weights
         ori_val = ori_weights[1].numpy()
@@ -328,8 +354,11 @@ def test_lambda_layer_keras_layer(self):
         output1 = M1(npInput).numpy()
         output3 = M3(npInput).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M3_config = RemoveDateInConfig(M3.config)
+
         self.assertEqual((output1 == output3).all(), True)
-        self.assertEqual(M1.config, M3.config)
+        self.assertEqual(M1_config, M3_config)
 
         ori_weights = M3.all_weights
         ori_val = ori_weights[1].numpy()
@@ -365,8 +394,11 @@ def func(noise, mean, std, foo=42):
         output1 = M1(ipt).numpy()
         output2 = M2(ipt).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
         self.assertEqual((output1 == output2).all(), True)
-        self.assertEqual(M1.config, M2.config)
+        self.assertEqual(M1_config, M2_config)
 
     def test_elementwise_no_para_no_args(self):
         # z = mean + noise * tf.exp(std * 0.5) + foo
@@ -387,8 +419,11 @@ def func(noise, mean, std, foo=42):
         output1 = M1(ipt).numpy()
         output2 = M2(ipt).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
         self.assertEqual((output1 == output2).all(), True)
-        self.assertEqual(M1.config, M2.config)
+        self.assertEqual(M1_config, M2_config)
 
     def test_elementwise_lambda_func(self):
         # z = mean + noise * tf.exp(std * 0.5)
@@ -410,8 +445,11 @@ def test_elementwise_lambda_func(self):
         output1 = M1(ipt).numpy()
         output2 = M2(ipt).numpy()
 
+        M1_config = RemoveDateInConfig(M1.config)
+        M2_config = RemoveDateInConfig(M2.config)
+
         self.assertEqual((output1 == output2).all(), True)
-        self.assertEqual(M1.config, M2.config)
+        self.assertEqual(M1_config, M2_config)
 
     # # ElementwiseLambda does not support keras layer/model func yet
     # def test_elementwise_keras_model(self):
@@ -434,8 +472,11 @@ def test_elementwise_lambda_func(self):
     #     output1 = M1(ipt).numpy()
     #     output2 = M2(ipt).numpy()
     #
+    #     M1_config = RemoveDateInConfig(M1.config)
+    #     M2_config = RemoveDateInConfig(M2.config)
+    #
     #     self.assertEqual((output1 == output2).all(), True)
-    #     self.assertEqual(M1.config, M2.config)
+    #     self.assertEqual(M1_config, M2_config)
 
 
 class basic_dynamic_model(Model):