run.py 6.79 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
"""Brain Mapper Run File

Description:
-------------
This file contains all the relevant functions for running BrainMapper.
The network can be ran in one of these modes:
    - train
    - evaluate path
    - evaluate whole


Usage
-------------
In order to run the network, in the terminal, the user needs to pass it relevant arguments:
    - (TODO: ADD ARGUMENTS)

"""

19
20
import torch
from utils.data_utils import get_datasets
Andrei-Claudiu Roibu's avatar
Andrei-Claudiu Roibu committed
21
from BrainMapperUNet import BrainMapperUNet
22
import torch.utils.data as data
23
24
from solver import Solver
import os
25
26
27
28
29
30
31
32
33
34
35
36
37
38

# Set the default floating point tensor type to FloatTensor

torch.set_default_tensor_type(torch.FloatTensor)

def load_data(data_parameters):
    """Dataset Loader

    This function loads the training and testing datasets.
    TODO: Will need to define if all the training data is loaded as bulk or individually!

    Args:
        data_parameters (dict): Dictionary containing relevant information for the datafiles.
        data_parameters = {
39
40
41
42
43
            'data_directory': 'path/to/directory'
            'train_data_file': 'training_data'
            'train_output_targets': 'training_targets'
            'test_data_file': 'testing_data'
            'test_target_file': 'testing_targets'
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
        }

    Returns:
        train_data (dataset object): Pytorch map-style dataset object, mapping indices to training data samples.
        test_data (dataset object): Pytorch map-style dataset object, mapping indices to testing data samples.

    Raises:
        None

    """
    print("Data is loading...")
    train_data, test_data = get_datasets(data_parameters)
    print("Data has loaded!")
    print("Training dataset size is {}".format(len(train_data)))
    print("Testing dataset size is {}".format(len(test_data)))

    return train_data, test_data
61

62
63
def train(data_parameters, training_parameters, network_parameters, misc_parameters):
    """Training Function
64
    
65
    This function trains a given model using the provided training data.
66
67
68
69
70
71
    Currently, the data loaded is set to have multiple sub-processes. 
    A high enough number of workers assures that CPU computations are efficiently managed, i.e. that the bottleneck is indeed the neural network's forward and backward operations on the GPU (and not data generation)
    Loader memory is also pinned, to speed up data transfer from CPU to GPU  by using the page-locked memory.
    Train data is also re-shuffled at each training epoch. 

    Args:
72
73
74
75
76
77
78
        data_parameters (dict): Dictionary containing relevant information for the datafiles.
        data_parameters = {
            'data_directory': 'path/to/directory'
            'train_data_file': 'training_data'
            'train_output_targets': 'training_targets'
            'test_data_file': 'testing_data'
            'test_target_file': 'testing_targets'
Andrei-Claudiu Roibu's avatar
Andrei-Claudiu Roibu committed
79
        }
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97

        training_parameters(dict): Dictionary containing relevant hyperparameters for training the network.
        training_parameters = {
            'training_batch_size': 5
            'test_batch_size: 5
            'use_pre_trained': False
            'pre_trained_path': 'pre_trained/path'
            'experiment_name': 'experiment_name'
            'learning_rate': 1e-4
            'optimizer_beta': (0.9, 0.999)
            'optimizer_epsilon': 1e-8
            'optimizer_weigth_decay': 1e-5
            'number_of_epochs': 10
            'loss_log_period': 50
            'learning_rate_scheduler_step_size': 3
            'learning_rate_scheduler_gamma': 1e-1
            'use_last_checkpoint': True
            'final_model_output_file': 'path/to/model'
98
        }
99
100
101

        network_parameters (dict): Contains information relevant parameters 
        network_parameters= {
102
103
104
105
106
107
108
109
110
111
112
113
114
            'kernel_heigth': 5
            'kernel_width': 5
            'kernel_classification': 1
            'input_channels': 1
            'output_channels': 64
            'convolution_stride': 1
            'dropout': 0.2
            'pool_kernel_size': 2
            'pool_stride': 2
            'up_mode': 'upconv'
            'number_of_classes': 1
        }

115
116
117
118
119
120
121
122
123
        misc_parameters (dict): Dictionary of aditional hyperparameters
        misc_parameters = {
            'save_model_directory': 'directory_name'
            'model_name': 'BrainMapper'
            'logs_directory': 'log-directory'
            'device': 1
            'experiments_directory': 'experiments-directory'
        }

124
125
126
127
128
129
130
131
132
133
134
    Returns:
        None

    Raises:
        None
    """

    train_data, test_data = load_data(data_parameters)

    train_loader = data.DataLoader(
        dataset= train_data,
135
        batch_size= training_parameters['training_batch_size'],
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
        shuffle= True,
        num_workers= 4,
        pin_memory= True
    )

    test_loader = data.DataLoader(
        dataset= test_data,
        batch_size= training_parameters['test_batch_size'],
        shuffle= False,
        num_workers= 4,
        pin_memory= True 
    )

    if training_parameters['use_pre_trained']:
        BrainMapperModel = torch.load(training_parameters['pre_trained_path'])
    else:
        BrainMapperModel = BrainMapperUNet(network_parameters)

154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
    solver = Solver(model= BrainMapperModel,
                    device= misc_parameters['device'],
                    number_of_classes= network_parameters['number_of_classes'],
                    experiment_name= training_parameters['experiment_name'],
                    optimizer_arguments = {'lr': training_parameters['learning_rate'],
                                            'betas': training_parameters['optimizer_beta'],
                                            'eps': training_parameters['optimizer_epsilon'],
                                            'weight_decay': training_parameters['optimizer_weigth_decay']
                                            },
                    model_name = misc_parameters['model_name'],
                    number_epochs = training_parameters['number_of_epochs'],
                    loss_log_period = training_parameters['loss_log_period'],
                    learning_rate_scheduler_step_size = training_parameters['learning_rate_scheduler_step_size'],
                    learning_rate_scheduler_gamma = training_parameters['learning_rate_scheduler_gamma'],
                    use_last_checkpoint = training_parameters['use_last_checkpoint'],
                    experiment_directory = misc_parameters['experiments_directory'],
                    logs_directory = misc_parameters['logs_directory']
                    )

    solver.train(train_loader, test_loader)

    model_output_path = os.path.join(misc_parameters['save_model_directory'], training_parameters['final_model_output_file'])
    BrainMapperModel.save(model_output_path)

    print("Final Model Saved in: {}".format(model_output_path))
179
180
181
182
183
184
185
186
187
188
189
190

def evaluate_path():
    pass

def evaluate_network():
    pass

def delete_files():
    pass

if __name__ == '__main__':
    pass