Fix model load .

Author: oaksharks

deeptables/models/deepmodel.py

@@ -28,8 +28,8 @@ def __init__(self,
                  config,
                  categorical_columns,
                  continuous_columns,
-                 var_categorical_len_columns=None,  # Compatible persisted model
-                 model_file=None):
+                 model_file=None,
+                 var_categorical_len_columns=None, ):
 
         # set gpu usage strategy before build model
         if config.gpu_usage_strategy == consts.GPU_USAGE_STRATEGY_GROWTH:

deeptables/models/deeptable.py

@@ -336,6 +336,8 @@ def fit(self, X=None, y=None, batch_size=128, epochs=1, verbose=1, callbacks=Non
             max_queue_size=10, workers=1, use_multiprocessing=False):
         logger.info(f'X.Shape={np.shape(X)}, y.Shape={np.shape(y)}, batch_size={batch_size}, config={self.config}')
         logger.info(f'metrics:{self.config.metrics}')
+        if np.ndim(X) != 2:
+            raise ValueError("Input train data should be 2d .")
         X_shape = np.shape(X)
         if X_shape[1] < 1:
             raise ValueError("Input train data should has 1 feature at least.")
@@ -353,7 +355,7 @@ def fit(self, X=None, y=None, batch_size=128, epochs=1, verbose=1, callbacks=Non
         model = DeepModel(self.task, self.num_classes, self.config,
                           self.preprocessor.categorical_columns,
                           self.preprocessor.continuous_columns,
-                          self.preprocessor.var_len_categorical_columns)
+                          var_categorical_len_columns=self.preprocessor.var_len_categorical_columns)
         history = model.fit(X, y, batch_size=batch_size, epochs=epochs, verbose=verbose, shuffle=shuffle,
                             validation_split=validation_split, validation_data=validation_data,
                             validation_steps=validation_steps, validation_freq=validation_freq,
@@ -706,7 +708,8 @@ def load_deepmodel(self, filepath):
         if os.path.exists(filepath):
             print(f'Load model from disk:{filepath}.')
             dm = DeepModel(self.task, self.num_classes, self.config,
-                           self.preprocessor.categorical_columns, self.preprocessor.continuous_columns, filepath)
+                           self.preprocessor.categorical_columns,
+                           self.preprocessor.continuous_columns, model_file=filepath)
             return dm
         else:
             raise ValueError(f'Invalid model filename:{filepath}.')