Tutorial

This tutorial would walk through the basic usage of using USB to benchmark any provided algorithms, customize algorithms, customize dataset and the introduction of lighting version.

Get Started

---

Lighting in 15 mins

---

This tutorial will walk you through the basics of using the torchssl-lightning framework. Let’s get started by training a FixMatch model on CIFAR-10!

import sys
sys.path.append('../')
from src import get_dataset, get_data_loader, net_builder, get_algorithm, get_config, Trainer

---

1. Define Configs and Create Config

config = {
    'algorithm': 'fixmatch',
    'net': 'wrn_28_2',
    'use_pretrain': False,  # todo: add pretrain
    'pretrain_path': None,

    # optimization configs
    'epoch': 3,
    'num_train_iter': 150,
    'num_eval_iter': 50,
    'optim': 'SGD',
    'lr': 0.03,
    'momentum': 0.9,
    'batch_size': 64,
    'eval_batch_size': 64,

    # dataset configs
    'dataset': 'cifar10',
    'num_labels': 40,
    'num_classes': 10,
    'input_size': 32,
    'data_dir': './data',

    # algorithm specific configs
    'hard_label': True,
    'uratio': 3,
    'ulb_loss_ratio': 1.0,

    # device configs
    'gpu': 0,
    'world_size': 1,
    'distributed': False,
}
config = get_config(config)

---

2. Create Model and Specify Algorithm

algorithm = get_algorithm(config,  net_builder(config.net, from_name=False), tb_log=None, logger=None)

---

3. Create Dataset and Dataloader

# create dataset
dataset_dict = get_dataset(config, config.algorithm, config.dataset, config.num_labels,
                           config.num_classes, data_dir=config.data_dir)
# create data loader for labeled training set
train_lb_loader = get_data_loader(config, dataset_dict['train_lb'], config.batch_size)
# create data loader for unlabeled training set
train_ulb_loader = get_data_loader(config, dataset_dict['train_ulb'], int(config.batch_size * config.uratio))
# create data loader for evaluation
eval_loader = get_data_loader(config, dataset_dict['eval'], config.eval_batch_size)

---

4. Train

trainer = Trainer(config, algorithm)
trainer.fit(train_lb_loader, train_ulb_loader, eval_loader)

---

5. Evaluate

trainer.evaluate(eval_loader)

---

6. Predict

y_pred, y_logits = trainer.predict(eval_loader)

Customized Usage

---

Customizing Dataset

---

In this tutorial, we provide an example of using custom dataset.

1. Specifiy configs and define the model

# define configs and create config
config = {
    'algorithm': 'fixmatch',
    'net': 'wrn_28_2',
    'use_pretrain': False,  # todo: add pretrain

    # optimization configs
    'epoch': 3,
    'num_train_iter': 150,
    'num_eval_iter': 50,
    'optim': 'SGD',
    'lr': 0.03,
    'momentum': 0.9,
    'batch_size': 64,
    'eval_batch_size': 64,

    # dataset configs
    'dataset': 'none',
    'num_labels': 40,
    'num_classes': 10,
    'input_size': 32,
    'data_dir': './data',

    # algorithm specific configs
    'hard_label': True,
    'uratio': 3,
    'ulb_loss_ratio': 1.0,

    # device configs
    'gpu': 0,
    'world_size': 1,
    'distributed': False,
}
config = get_config(config)

# create model and specify algorithm
algorithm = get_algorithm(config,  net_builder(config.net, from_name=False), tb_log=None, logger=None)

---

2. Create Dataset

# replace with your own code
data = np.random.randint(0, 255, size=3072 * 1000).reshape((-1, 32, 32, 3))
data = np.uint8(data)
target = np.random.randint(0, 10, size=1000)
lb_data, lb_target, ulb_data, ulb_target = split_ssl_data(config, data, target,
                                                          num_labels=config.num_labels,
                                                          num_classes=config.num_classes)

train_transform = transforms.Compose([transforms.RandomHorizontalFlip(),
                                      transforms.RandomCrop(32, padding=int(32 * 0.125), padding_mode='reflect'),
                                      transforms.ToTensor(),
                                      transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])])

lb_dataset = BasicDataset(config.algorithm, lb_data, lb_target, config.num_classes, train_transform, is_ulb=False)
ulb_dataset = BasicDataset(config.algorithm, lb_data, lb_target, config.num_classes, train_transform, is_ulb=True)

# replace with your own code
eval_data = np.random.randint(0, 255, size=3072 * 100).reshape((-1, 32, 32, 3))
eval_data = np.uint8(eval_data)
eval_target = np.random.randint(0, 10, size=100)

eval_transform = transforms.Compose([transforms.Resize(32),
                                      transforms.ToTensor(),
                                      transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])])

eval_dataset = BasicDataset(config.algorithm, lb_data, lb_target, config.num_classes, eval_transform, is_ulb=False)

# define data loaders
train_lb_loader = get_data_loader(config, lb_dataset, config.batch_size)
train_ulb_loader = get_data_loader(config, ulb_dataset, int(config.batch_size * config.uratio))
eval_loader = get_data_loader(config, eval_dataset, config.eval_batch_size)

---

3. Training and Evaluation

# training and evaluation
trainer = Trainer(config, algorithm)
trainer.fit(train_lb_loader, train_ulb_loader, eval_loader)
trainer.evaluate(eval_loader)