Exploiting the Relationship between Pruning Ratio and Compression Effect for Neural Network Model Based on TensorFlow

Table 2

Descriptions of parameters of the pruning effect view command (view effects of −1 or −2 training).

Parameter	Description	Default value
--test_data_dir	Folder path containing test data	/tmp/mnist_data
--checkpoint_dir	Folder path of the trained model	/tmp/mnist_train/step_2_2
--variables_dir	Folder containing data of the last trained model	/tmp/mnist_variables
--max_steps	Number of iterations during model testing	10
--batch_size	Size of sample data of each batch used for calculation during model testing	32