add documentation for CNN, allow it to interact with model_training_parameters
This commit is contained in:
@@ -17,8 +17,6 @@ logger = logging.getLogger(__name__)
|
||||
# tf.config.run_functions_eagerly(True)
|
||||
# tf.data.experimental.enable_debug_mode()
|
||||
|
||||
MAX_EPOCHS = 10
|
||||
|
||||
|
||||
class CNNPredictionModel(BaseTensorFlowModel):
|
||||
"""
|
||||
@@ -46,7 +44,12 @@ class CNNPredictionModel(BaseTensorFlowModel):
|
||||
)
|
||||
|
||||
n_features = len(data_dictionary["train_features"].columns)
|
||||
BATCH_SIZE = self.freqai_info.get("batch_size", 64)
|
||||
BATCH_SIZE = self.model_training_parameters.get("batch_size", 64)
|
||||
|
||||
# we need to remove batch_size from the model_training_params because
|
||||
# we dont want fit() to get the incorrect assignment (we use the WindowGenerator)
|
||||
# to handle our batches.
|
||||
self.model_training_parameters.pop('batch_size')
|
||||
input_dims = [BATCH_SIZE, self.CONV_WIDTH, n_features]
|
||||
|
||||
w1 = WindowGenerator(
|
||||
@@ -84,8 +87,7 @@ class CNNPredictionModel(BaseTensorFlowModel):
|
||||
|
||||
model.fit(
|
||||
w1.train,
|
||||
epochs=MAX_EPOCHS,
|
||||
shuffle=False,
|
||||
**self.model_training_parameters,
|
||||
validation_data=val_data,
|
||||
callbacks=[early_stopping],
|
||||
verbose=1,
|
||||
|
Reference in New Issue
Block a user