Skip to content

Vision Experiments #104

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xinthelian wants to merge 1 commit into
base: torch_vision
Choose a base branch
from
Open

Vision Experiments #104

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
963 changes: 963 additions & 0 deletions concept_formation/vision-experiments/datasets_cifar.py
View file
Open in desktop

Large diffs are not rendered by default.

664 changes: 664 additions & 0 deletions concept_formation/vision-experiments/datasets_mnist.py
View file
Open in desktop

Large diffs are not rendered by default.

115 changes: 115 additions & 0 deletions concept_formation/vision-experiments/experiment_0.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,115 @@
import torch

from concept_formation.cobweb_torch import CobwebTorchTree
from datasets_cifar import dataloaders_0
import models_nn
import models_cobweb


def reconfig(general_config, model_config, data_config):
"""
Re-initialization for model_config and data_config based on requirements of experiments 0.
"""
if data_config['dataset'] == 'mnist':
data_config['split_size'] = 6000
else:
data_config['split_size'] = 5000
if model_config['type'] == 'cobweb':
general_config['cuda'] == False
data_config['batch_size_tr'] = 60000
data_config['batch_size_te'] = 10000


def experiment_0_nn(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)

# Dataloaders:
dataloaders = dataloaders_0(general_config, data_config, dataset_tr, dataset_te, verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
device = torch.device("cuda" if general_config['cuda'] else "cpu")
model, optimizer = models_nn.build_model(model_config, data_config, device)

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 0")
print("Experiments description: Train data from all labels with sequential splits.")
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # fc or cnn
print("Seed:", general_config['seed'])
print("Epochs:", model_config['epoch'])
print("\nModel overview:")
print(model)
print("\nOptimizer:")
print(optimizer)
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))

for epoch in range(1, model_config['epoch'] + 1):
if verbose:
print("\n\n [Epoch {}]".format(epoch))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_nn.train(model, optimizer, loaders_tr[i], epoch, model_config['log_interval'], device)

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_nn.test(model, loaders_te[j], device)
test_accs.append(acc.item())

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs


def experiment_0_cobweb(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)

# Dataloaders:
dataloaders = dataloaders_0(general_config, data_config, dataset_tr, dataset_te, verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
example_imgs, _ = next(iter(loaders_tr[0]))
model = CobwebTorchTree(example_imgs.shape[1:])

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 0")
print("Experiments description: Train data from all labels with sequential splits.")
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # cobweb
print("Seed:", general_config['seed'])
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_cobweb.train(model, loaders_tr[i])

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_cobweb.test(model, loaders_te[j])
print("Test accuracy: {}".format(acc))
test_accs.append(acc)

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs

118 changes: 118 additions & 0 deletions concept_formation/vision-experiments/experiment_1.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,118 @@
import torch

from concept_formation.cobweb_torch import CobwebTorchTree
from datasets_mnist import dataloaders_1
import models_nn
import models_cobweb


def reconfig(general_config, model_config, data_config):
"""
Re-initialization for model_config and data_config based on requirements of experiments 0.
"""
if data_config['dataset'] == 'mnist':
data_config['split_size'] = 6000
else:
data_config['split_size'] = 5000
if model_config['type'] == 'cobweb':
general_config['cuda'] == False
data_config['batch_size_tr'] = 60000
data_config['batch_size_te'] = 10000


def experiment_1_nn(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)
label = general_config['label']

# Dataloaders:
dataloaders = dataloaders_1(general_config, data_config, dataset_tr, dataset_te, verbose=verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
device = torch.device("cuda" if general_config['cuda'] else "cpu")
model, optimizer = models_nn.build_model(model_config, data_config, device)

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 1")
print("Experiments description: Train all data from label {} first, then train the rest with sequential splits.".format(label))
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # fc or cnn
print("Seed:", general_config['seed'])
print("The selected label:", label)
print("Epochs:", model_config['epoch'])
print("\nModel overview:")
print(model)
print("\nOptimizer:")
print(optimizer)
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))

for epoch in range(1, model_config['epoch'] + 1):
if verbose:
print("\n\n [Epoch {}]".format(epoch))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_nn.train(model, optimizer, loaders_tr[i], epoch, model_config['log_interval'], device)

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_nn.test(model, loaders_te[j], device)
test_accs.append(acc.item())

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs


def experiment_1_cobweb(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)
label = general_config['label']

# Dataloaders:
dataloaders = dataloaders_1(general_config, data_config, dataset_tr, dataset_te, verbose=verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
example_imgs, _ = next(iter(loaders_tr[0]))
model = CobwebTorchTree(example_imgs.shape[1:])

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 1")
print("Experiments description: Train all data from label {} first, then train the rest with sequential splits.".format(label))
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # cobweb
print("Seed:", general_config['seed'])
print("The selected label:", label)
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_cobweb.train(model, loaders_tr[i])

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_cobweb.test(model, loaders_te[j])
print("Test accuracy: {}".format(acc))
test_accs.append(acc)

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs
124 changes: 124 additions & 0 deletions concept_formation/vision-experiments/experiment_2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,124 @@
import torch

from concept_formation.cobweb_torch import CobwebTorchTree
from datasets_mnist import dataloaders_2
import models_nn
import models_cobweb


def reconfig(general_config, model_config, data_config):
"""
Re-initialization for model_config and data_config based on requirements of experiments 0.
"""
if data_config['dataset'] == 'mnist':
data_config['split_size'] = 5400
data_config['size_all_tr_each'] = 600
else:
data_config['split_size'] = 4500
data_config['size_all_tr_each'] = 500
if model_config['type'] == 'cobweb':
general_config['cuda'] == False
data_config['batch_size_tr'] = 60000
data_config['batch_size_te'] = 10000


def experiment_2_nn(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)
label = general_config['label']
size_all_tr_each = data_config['size_all_tr_each']

# Dataloaders:
dataloaders = dataloaders_2(general_config, data_config, dataset_tr, dataset_te, verbose=verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
device = torch.device("cuda" if general_config['cuda'] else "cpu")
model, optimizer = models_nn.build_model(model_config, data_config, device)

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 2")
print("Experiments description: Train all data from label {} and some data ({}) from each remaining label first,"
" then train the rest with sequential splits.".format(label, size_all_tr_each))
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # fc or cnn
print("Seed:", general_config['seed'])
print("The selected label:", label)
print("Epochs:", model_config['epoch'])
print("\nModel overview:")
print(model)
print("\nOptimizer:")
print(optimizer)
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))

for epoch in range(1, model_config['epoch'] + 1):
if verbose:
print("\n\n [Epoch {}]".format(epoch))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_nn.train(model, optimizer, loaders_tr[i], epoch, model_config['log_interval'], device)

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_nn.test(model, loaders_te[j], device)
test_accs.append(acc.item())

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs


def experiment_2_cobweb(general_config, model_config, data_config, dataset_tr, dataset_te, verbose):

reconfig(general_config, model_config, data_config)
label = general_config['label']
size_all_tr_each = data_config['size_all_tr_each']

# Dataloaders:
dataloaders = dataloaders_2(general_config, data_config, dataset_tr, dataset_te, verbose=verbose)
loaders_tr = dataloaders.training_loaders
loaders_te = dataloaders.test_loaders

# Models and optimizers:
example_imgs, _ = next(iter(loaders_tr[0]))
model = CobwebTorchTree(example_imgs.shape[1:])

# Store the test accuracies
test_accs = []

if verbose:
print('\n\n' + ' START EXPERIMENTS '.center(70, '~'))
print("Experiments type: 2")
print("Experiments description: Train all data from label {} and some data ({}) from each remaining label first,"
" then train the rest with sequential splits.".format(label, size_all_tr_each))
print("Number of Train-test trials:", len(loaders_tr))
print("Model:", model_config['type']) # cobweb
print("Seed:", general_config['seed'])
print("The selected label:", label)
print("\nCUDA is {}used.".format("" if general_config['cuda'] else "NOT "))

for i in range(len(loaders_tr)):
if verbose:
print("\n\n" + " Trial {} ".format(i+1).center(70, '='))
print("\n====> Model Training with labels {} <====".format(dataloaders.labels_tr[i]))
models_cobweb.train(model, loaders_tr[i])

for j in range(len(loaders_te)):
if verbose:
print("\n----> Model Testing with labels {} <----".format(dataloaders.labels_te[j]))
acc = models_cobweb.test(model, loaders_te[j])
print("Test accuracy: {}".format(acc))
test_accs.append(acc)

print("\n\nThis is the end of the experiments.")
print("There are {} test accuracy data in total.".format(len(test_accs)))
return test_accs
Loading