def test_batch_norm_class(self):
# This tests the model and trainer set up
train_config_text_proto = """
optimizer {
gradient_descent {
learning_rate {
constant_learning_rate {
learning_rate: 1.0
}
}
}
}
max_iterations: 5
"""
model_config_text_proto = """
path_drop_probabilities: [1.0, 1.0]
"""
train_config = train_pb2.TrainConfig()
text_format.Merge(train_config_text_proto, train_config)
model_config = model_pb2.ModelConfig()
text_format.Merge(model_config_text_proto, model_config)
train_config.overwrite_checkpoints = True
test_root_dir = '/tmp/avod_unit_test/'
paths_config = model_config.paths_config
paths_config.logdir = test_root_dir + 'logs/'
paths_config.checkpoint_dir = test_root_dir
classifier = FakeBatchNormClassifier(model_config)
trainer.train(classifier, train_config)
def get_model_config_from_file(config_path):
"""Reads model configuration from a configuration file.
This merges the layer config info with model default configs.
Args:
config_path: A path to the config
Returns:
layers_config: A configured model_pb2 config
"""
model_config = model_pb2.ModelConfig()
with open(config_path, 'r') as f:
text_format.Merge(f.read(), model_config)
return model_config