def save(self, name=None, model_plot=False):
"""
Save the model to disk
:param name: Folder name to be saved
:type name: string
:param model_plot: True to plot model too
:type model_plot: boolean
:return: A saved folder on disk
"""
# Only generate a folder automatically if no name provided
if self.folder_name is None and name is None:
self.folder_name = folder_runnum()
else:
if name is not None:
self.folder_name = name
# if foldername provided, then create a directory
if not os.path.exists(os.path.join(self.currentdir, self.folder_name)):
os.makedirs(os.path.join(self.currentdir, self.folder_name))
self.fullfilepath = os.path.join(self.currentdir,
self.folder_name + '/')
txt_file_path = self.fullfilepath + 'hyperparameter.txt'
if os.path.isfile(txt_file_path):
self.hyper_txt = open(txt_file_path, 'a')
self.hyper_txt.write("\n")
self.hyper_txt.write("======Another Run======")
else:
self.hyper_txt = open(txt_file_path, 'w')
self.hyper_txt.write(f"Model: {self.name} \n")
self.hyper_txt.write(f"Model Type: {self._model_type} \n")
self.hyper_txt.write(
f"astroNN identifier: {self._model_identifier} \n")
self.hyper_txt.write(f"Python Version: {self._python_info} \n")
self.hyper_txt.write(f"astroNN Version: {self._astronn_ver} \n")
self.hyper_txt.write(f"Keras Version: {self._keras_ver} \n")
self.hyper_txt.write(f"Tensorflow Version: {self._tf_ver} \n")
self.hyper_txt.write(f"Folder Name: {self.folder_name} \n")
self.hyper_txt.write(f"Batch size: {self.batch_size} \n")
self.hyper_txt.write(
f"Optimizer: {self.optimizer.__class__.__name__} \n")
self.hyper_txt.write(f"Maximum Epochs: {self.max_epochs} \n")
self.hyper_txt.write(f"Learning Rate: {self.lr} \n")
self.hyper_txt.write(f"Validation Size: {self.val_size} \n")
self.hyper_txt.write(f"Input Shape: {self.input_shape} \n")
self.hyper_txt.write(f"Label Shape: {self.labels_shape} \n")
self.hyper_txt.write(f"Number of Training Data: {self.num_train} \n")
self.hyper_txt.write(f"Number of Validation Data: {self.val_num} \n")
if model_plot is True:
self.plot_model()
self.post_training_checklist_child()
self.virtual_cvslogger.savefile(folder_name=self.folder_name)
def save(self, name=None, model_plot=False):
"""
Save the model to disk
:param name: Folder name to be saved
:type name: string
:param model_plot: True to plot model too
:type model_plot: boolean
:return: A saved folder on disk
"""
self.has_model_check()
# Only generate a folder automatically if no name provided
if self.folder_name is None and name is None:
self.folder_name = folder_runnum()
elif name is not None:
self.folder_name = name
# if foldername provided, then create a directory, if exist append something to avoid overwrite
if not os.path.exists(os.path.join(self.currentdir, self.folder_name)):
os.makedirs(os.path.join(self.currentdir, self.folder_name))
else:
i_back = 2
while True:
if not os.path.exists(
os.path.join(self.currentdir,
self.folder_name + f'_{i_back}')):
break
i_back += 1
new_folder_name_temp = self.folder_name + f'_{i_back}'
warnings.warn(
f'To prevent your model being overwritten, your folder name changed from {self.folder_name} '
f'to {new_folder_name_temp}', UserWarning)
self.folder_name = new_folder_name_temp
os.makedirs(os.path.join(self.currentdir, self.folder_name))
self.fullfilepath = os.path.join(self.currentdir,
self.folder_name + os.sep)
txt_file_path = self.fullfilepath + 'hyperparameter.txt'
if os.path.isfile(txt_file_path):
self.hyper_txt = open(txt_file_path, 'a')
self.hyper_txt.write("\n")
self.hyper_txt.write("======Another Run======")
else:
self.hyper_txt = open(txt_file_path, 'w')
self.hyper_txt.write(f"Model: {self.name} \n")
self.hyper_txt.write(f"Model Type: {self._model_type} \n")
self.hyper_txt.write(
f"astroNN identifier: {self._model_identifier} \n")
self.hyper_txt.write(f"Python Version: {self._python_info} \n")
self.hyper_txt.write(f"astroNN Version: {self._astronn_ver} \n")
self.hyper_txt.write(f"Keras Version: {self._keras_ver} \n")
self.hyper_txt.write(f"Tensorflow Version: {self._tf_ver} \n")
self.hyper_txt.write(f"Folder Name: {self.folder_name} \n")
self.hyper_txt.write(f"Batch size: {self.batch_size} \n")
self.hyper_txt.write(
f"Optimizer: {self.optimizer.__class__.__name__} \n")
self.hyper_txt.write(f"Maximum Epochs: {self.max_epochs} \n")
self.hyper_txt.write(f"Learning Rate: {self.lr} \n")
self.hyper_txt.write(f"Validation Size: {self.val_size} \n")
self.hyper_txt.write(f"Input Shape: {self._input_shape} \n")
self.hyper_txt.write(f"Label Shape: {self._labels_shape} \n")
self.hyper_txt.write(f"Number of Training Data: {self.num_train} \n")
self.hyper_txt.write(f"Number of Validation Data: {self.val_num} \n")
if model_plot is True:
self.plot_model()
self.post_training_checklist_child()
if self.virtual_cvslogger is not None: # in case you save without training, so cvslogger is None
self.virtual_cvslogger.savefile(folder_name=self.folder_name)