def run_cifar_user_input_bidi():
status = program.Status()
status.max_layer_conv2d = int(
input("Enter max convolution layers model: "))
status.max_filter = 530
status.max_filter_dense = 270
status.max_kernel_dense = 9
status.max_pool_layer = 4
status.max_parents = 2
list_conditions = {
DNA_conditions.max_filter: status.max_filter,
DNA_conditions.max_filter_dense: status.max_filter_dense,
DNA_conditions.max_kernel_dense: status.max_kernel_dense,
DNA_conditions.max_layer: status.max_layer_conv2d,
DNA_conditions.min_filter: 3,
DNA_conditions.max_pool_layer: status.max_pool_layer,
DNA_conditions.max_parents: status.max_parents,
DNA_conditions.no_con_last_layer: 1,
}
def condition(DNA):
return DNA_conditions.dict2condition(DNA, list_conditions)
def dropout_function(base_p, total_layers, index_layer, isPool=False):
value = 0
if index_layer != 0 and isPool == False:
value = 0.05
if index_layer == total_layers - 2:
value = 0.05
return value
settings = ExperimentSettings.ExperimentSettings()
settings.version = directions_version.CONVEX_VERSION
settings.dropout_function = dropout_function
settings.eps_batchorm = 0.001
settings.momentum = 0.9
settings.enable_activation = True
settings.enable_last_activation = True
settings.enable_augmentation = True
settings.enable_track_stats = True
settings.dropout_value = 0.05
settings.weight_decay = 0.0005
settings.evalLoss = True
# Batch size
status.S = 64
e = 50000 / status.S
status.iterations_per_epoch = math.ceil(e)
status.condition = condition
status.dt_Max = 0.05
status.dt_min = 0.0000001
status.clear_period = 200000
status.max_iter = 400000
status.restart_period = 18 * status.iterations_per_epoch
status.max_layer = 8
#status.max_filter=51
from utilities.Abstract_classes.classes.uniform_random_selector_2 import (
centered_random_selector as Selector)
status.mutations = ((1, 0, 0, 0), (1, 0, 0, 0), (0, 1, 0, 0), (0, 1, 0, 0),
(4, 0, 0, 0), (0, 0, 1), (0, 0, -1), (0, 0, 1, 1),
(0, 0, -1, -1), (0, 0, 2))
status.num_actions = int(input("num_actions : "))
status.Selector_creator = Selector
status.log_size = 200
status.min_log_size = 100
status.version = directions_version.CONVEX_VERSION
status.cuda = True
augSettings = AugmentationSettings.AugmentationSettings()
dict_transformations = {
augSettings.baseline_customRandomCrop: True,
augSettings.randomHorizontalFlip: True,
augSettings.randomErase_1: True
}
transform_compose = augSettings.generateTransformCompose(
dict_transformations, False)
settings.transformations_compose = transform_compose
settings.ricap = Augmentation_Utils.Ricap(beta=0.3)
settings.cuda = status.cuda
status.mutation_coefficient = float(input("mutation_coefficient : "))
status.save2database = False
# DNA BASE
status.Center = DNAs.DNA_base
status.settings = settings
program.run(status)
def run_cifar_user_input_bidi(test_name, mutations_actions,
const_direction_version):
import Geometric.Products.Product_f_cifar_save_2 as program
#import Product_f_cifar_save_nm as program
save = True #bool(input('Insert any input to save'))
status = program.Status()
status.max_layer_conv2d = int(input("Enter max conv2d layers: "))
status.max_filter = 530
status.max_filter_dense = 270
status.max_kernel_dense = 9
status.max_pool_layer = 4
status.max_parents = 2
list_conditions = {
DNA_conditions.max_filter: status.max_filter,
DNA_conditions.max_filter_dense: status.max_filter_dense,
DNA_conditions.max_kernel_dense: status.max_kernel_dense,
DNA_conditions.max_layer: status.max_layer_conv2d,
DNA_conditions.min_filter: 3,
DNA_conditions.max_pool_layer: status.max_pool_layer,
DNA_conditions.max_parents: status.max_parents,
DNA_conditions.no_con_last_layer: 1,
}
def condition(DNA):
return DNA_conditions.dict2condition(DNA, list_conditions)
def dropout_function(base_p, total_layers, index_layer, isPool=False):
value = 0
if index_layer != 0 and isPool == False:
value = 0.05
if index_layer == total_layers - 2:
value = 0.05
return value
settings = ExperimentSettings.ExperimentSettings()
settings.version = const_direction_version #directions_version.POOL_VERSION
settings.dropout_function = dropout_function
settings.eps_batchorm = 0.001
settings.momentum = 0.9
custom = False #bool(input('No input for defaults :'))
if custom == True:
ENABLE_ACTIVATIO = int(
input("Enable last layer activation? (1 = yes, 0 = no): "))
ENABLE_LAST_ACTIVATION = int(
input("Enable last layer activation? (1 = yes, 0 = no): "))
ENABLE_AUGMENTATION = int(
input("Enable Data augmentation? (1 = yes, 0 = no): "))
ENABLE_TRACK = int(
input("Enable tracking var/mean batchnorm? (1 = yes, 0 = no): "))
settings.dropout_value = float(input("dropout value: "))
settings.weight_decay = float(input('weight_decay: '))
else:
ENABLE_ACTIVATION = 1
ENABLE_LAST_ACTIVATION = 1
ENABLE_AUGMENTATION = 1
ENABLE_TRACK = 1
settings.dropout_value = 0.05
settings.weight_decay = 0.0005
value = True
if ENABLE_ACTIVATION == 0:
value = False
settings.enable_activation = value
# ENABLE_LAST_ACTIVATION, enable/disable last layer relu
value = False
if ENABLE_LAST_ACTIVATION == 1:
value = True
settings.enable_last_activation = value
# ENABLE_AUGMENTATION, enable/disable data augmentation
value = True
if ENABLE_AUGMENTATION == 0:
value = False
ENABLE_AUGMENTATION = value
settings.enable_augmentation = value
# ALLOW TRACK BATCHNORM
value = True
if ENABLE_TRACK == 0:
value = False
settings.enable_track_stats = value
status.S = 64
e = 50000 / status.S
status.iterations_per_epoch = math.ceil(e)
status.condition = condition
status.dt_Max = 0.05
status.dt_min = 0.0000001
status.clear_period = 200000
status.max_iter = 400000
status.restart_period = 18 * status.iterations_per_epoch
status.max_layer = 8
#status.max_filter=51
from utilities.Abstract_classes.classes.uniform_random_selector_2 import (
centered_random_selector as Selector)
status.mutations = mutations_actions
status.num_actions = int(input("num_actions : "))
status.Selector_creator = Selector
status.log_size = 200 #int(input("Log size : "))
status.min_log_size = 100
status.version = settings.version
status.cuda = True #bool(input("Any input for cuda : "))
settings.evalLoss = True #bool(input("Any input to activate EvalLoss : "))
'''
augSettings = AugmentationSettings.AugmentationSettings()
dict_transformations = {
augSettings.baseline_customRandomCrop : True,
augSettings.randomHorizontalFlip : True,
augSettings.randomErase_1 : True
}
transform_compose = augSettings.generateTransformCompose(dict_transformations, False)
settings.transformations_compose = transform_compose
'''
settings.ricap = Augmentation_Utils.Ricap(beta=0.3)
settings.cuda = status.cuda
status.mutation_coefficient = float(input("mutation_coefficient : "))
if save:
if test_name is not None:
status.experiment_name = test_name
else:
status.experiment_name = input("insert experiment name : ")
status.save_space_period = 200 #int(input("save_space_period : "))
status.save_net_period = 600 #int(input("save_space_net_period : "))
status.save2database = save
x = 32
y = 32
status.Center = DNAs.DNA_base
status.settings = settings
program.run(status)
import DAO.database.dao.TestModelDAO as TestModelDAO
import random
import os
import utilities.NetworkStorage as NetworkStorage
import utilities.MemoryManager as MemoryManager
from commands import CommandCreateDataGen
import utilities.ExperimentSettings as ExperimentSettings
testModelDAO = TestModelDAO.TestModelDAO()
memoryManager = MemoryManager.MemoryManager()
settings = ExperimentSettings.ExperimentSettings()
settings.version = "convex"
settings.batch_size = 64
settings.cuda = True
settings.eps_batchorm = 0.001
settings.dropout_value = 0.05
settings.weight_decay = 0.0005
settings.momentum = 0.9
settings.enable_activation = True
settings.enable_last_activation = True
settings.enable_track_stats = True
test_id = int(input("Enter id test: "))
max_alai_time = int(input("Enter max alai time: "))
test_models = testModelDAO.findByLimitAlai(idTest=test_id,
limit_alai_time=max_alai_time)
dataCreator = CommandCreateDataGen.CommandCreateDataGen(cuda=True)
dataCreator.execute(compression=2,
def TestMemoryManager():
epochs = 0.2
batch_size = 64
def dropout_function(base_p, total_layers, index_layer, isPool=False):
value = 0
if index_layer != 0 and isPool == False:
value = base_p
if index_layer == total_layers - 2:
value = base_p
print("conv2d: ", index_layer, " - dropout: ", value, " - isPool: ",
isPool)
return value
settings = ExperimentSettings.ExperimentSettings()
settings.momentum = 0.9
settings.dropout_value = 0.05
settings.weight_decay = 0.0005
settings.enable_activation = True
settings.enable_last_activation = True
settings.enable_track_stats = True
settings.version = directions_version.CONVEX_VERSION
settings.eps_batchorm = 0.001
settings.dropout_function = dropout_function
settings.ricap = Augmentation.Ricap(beta=0.3)
dataGen = GeneratorFromCIFAR.GeneratorFromCIFAR(2,
batch_size,
threads=0,
dataAugmentation=True)
dataGen.dataConv2d()
memoryManager = MemoryManager.MemoryManager()
mutation_manager = MutationManager.MutationManager(
directions_version=settings.version)
dna = test_DNAs.DNA_base
e = 50000 / batch_size
e = math.ceil(e)
print("e: ", e)
print("total iterations: ", int(epochs * e))
dt_array = Alaising(1.2, 99, int(epochs * e))
input("press to continue: before load network")
network = nw_dendrites.Network(
dna,
cuda_flag=True,
momentum=settings.momentum,
weight_decay=settings.weight_decay,
enable_activation=settings.enable_activation,
enable_track_stats=settings.enable_track_stats,
dropout_value=settings.dropout_value,
dropout_function=settings.dropout_function,
enable_last_activation=settings.enable_last_activation,
version=settings.version,
eps_batchnorm=settings.eps_batchorm)
input("press to continue: before training network")
network.training_custom_dt(dataGenerator=dataGen,
dt_array=dt_array,
ricap=settings.ricap,
evalLoss=True)
network.generate_accuracy(dataGen)
print("net acc: ", network.get_accuracy())
input("press to continue: before save network")
memoryManager.saveTempNetwork(network)
input("press to continue: after save network")
input("press to continue: before load temp network")
network_loaded = memoryManager.loadTempNetwork(dna, settings)
input("press to continue: after load temp network")
network_loaded.generate_accuracy(dataGen)
print("loaded acc: ", network_loaded.get_accuracy())
input("press to continue: before mutate network (add filters layer 1)")
dna_mutate = direction_dna.increase_filters(1, network_loaded.dna)
network_mutate = mutation_manager.execute_mutation(network_loaded,
dna_mutate)
input("press to continue: after mutate network")
input("press to continue: before delete old network")
memoryManager.deleteNetwork(network_loaded)
input("press to continue: after delete old network")
network_mutate.generate_accuracy(dataGen)
print("mutated acc: ", network_mutate.get_accuracy())
input("press to conitnue: before training mutate network")
network_mutate.training_custom_dt(dataGenerator=dataGen,
dt_array=dt_array,
ricap=settings.ricap,
evalLoss=True)
input("press to conitnue: after training mutate network")
network_mutate.generate_accuracy(dataGen)
print("mutate net acc: ", network_mutate.get_accuracy())
input("press to continue: before save network")
memoryManager.saveTempNetwork(network_mutate)
input("press to continue: after save network")
input("press to continue: before load network")
network_loaded = memoryManager.loadTempNetwork(dna_mutate, settings)
input("press to continue: after load network")
network_loaded.generate_accuracy(dataGen)
print("loaded acc: ", network_loaded.get_accuracy())
input("press to continue: before mutate network (add layer pool 1)")
dna_mutate_2 = direction_dna.add_pool_layer(1, network_loaded.dna)
network_mutate = mutation_manager.execute_mutation(network_loaded,
dna_mutate_2)
input("press to continue: after mutate network")
input("press to continue: before delete old network")
memoryManager.deleteNetwork(network_loaded)
input("press to continue: after delete old network")
network_mutate.generate_accuracy(dataGen)
print("mutated acc: ", network_mutate.get_accuracy())
input("press to conitnue: before training mutate network")
network_mutate.training_custom_dt(dataGenerator=dataGen,
dt_array=dt_array,
ricap=settings.ricap,
evalLoss=True)
input("press to conitnue: after training mutate network")
network_mutate.generate_accuracy(dataGen)
print("mutate net acc: ", network_mutate.get_accuracy())
input("press to continue: before save network")
memoryManager.saveTempNetwork(network_mutate)
input("press to continue: after save network")