ssl_tools.benchmarks.main_mix_style

Classes

CNN_HaEtAl_1D

CNN_HaEtAl_1D_Backbone

CNN_HaEtAl_2D

CNN_HaEtAl_2D_Backbone

ConvolutionalBlock

ExperimentArgs

ResNet1DBase

ResNet1D_8

ResNetBlock

ResNetSE1D_5

ResNetSE1D_8

ResNetSEBlock

SimpleClassificationNet2

SqueezeAndExcitation1D

_ResNet1D

Functions

_run_experiment_wrapper(experiment_args)

cli_main(experiment)

conv3x3(in_planes, out_planes[, stride, groups, dilation])

3x3 convolution with padding

conv3x3_dynamic(in_planes, out_planes[, stride, ...])

3x3 convolution with padding

main_loo()

pretty_print_experiment_args(args[, indent])

run_serial(experiments)

run_using_ray(experiments[, ray_address])

Module Contents

class ssl_tools.benchmarks.main_mix_style.CNN_HaEtAl_1D(input_shape=(1, 6, 60), num_classes=6, learning_rate=0.001)

Bases: SimpleClassificationNet2

Parameters:

  • input_shape (Tuple[int, int, int])

  • num_classes (int)

  • learning_rate (float)

_calculate_fc_input_features(backbone, input_shape)

Run a single forward pass with a random input to get the number of features after the convolutional layers.

Parameters

backbonetorch.nn.Module

The backbone of the network

input_shapeTuple[int, int, int]

The input shape of the network.

Returns

int

The number of features after the convolutional layers.

Parameters:

  • backbone (torch.nn.Module)

  • input_shape (Tuple[int, int, int])

Return type:

int

_create_backbone(input_shape)
Parameters:

input_shape (Tuple[int, int])

Return type:

torch.nn.Module

_create_fc(input_features, num_classes)
Parameters:

  • input_features (int)

  • num_classes (int)

Return type:

torch.nn.Module

class ssl_tools.benchmarks.main_mix_style.CNN_HaEtAl_1D_Backbone(input_channels=1)

Bases: torch.nn.Module

Parameters:

input_channels (int)

forward(x)
class ssl_tools.benchmarks.main_mix_style.CNN_HaEtAl_2D(pad_at=(3,), input_shape=(1, 6, 60), num_classes=6, learning_rate=0.001)

Bases: SimpleClassificationNet2

Parameters:

  • pad_at (List[int])

  • input_shape (Tuple[int, int, int])

  • num_classes (int)

  • learning_rate (float)

_calculate_fc_input_features(backbone, input_shape)

Run a single forward pass with a random input to get the number of features after the convolutional layers.

Parameters

backbonetorch.nn.Module

The backbone of the network

input_shapeTuple[int, int, int]

The input shape of the network.

Returns

int

The number of features after the convolutional layers.

Parameters:

  • backbone (torch.nn.Module)

  • input_shape (Tuple[int, int, int])

Return type:

int

_create_backbone(input_shape)
Parameters:

input_shape (Tuple[int, int])

Return type:

torch.nn.Module

_create_fc(input_features, num_classes)
Parameters:

  • input_features (int)

  • num_classes (int)

Return type:

torch.nn.Module

class ssl_tools.benchmarks.main_mix_style.CNN_HaEtAl_2D_Backbone(pad_at, in_channels=1)

Bases: torch.nn.Module

Parameters:

  • pad_at (int)

  • in_channels (int)

forward(x)
class ssl_tools.benchmarks.main_mix_style.ConvolutionalBlock(in_channels, activation_cls=None)

Bases: torch.nn.Module

Parameters:

  • in_channels (int)

  • activation_cls (torch.nn.Module)

forward(x)
class ssl_tools.benchmarks.main_mix_style.ExperimentArgs
data_cls: Any
mix: bool = True
model_args: Dict[str, Any]
model_cls: Any
seed: int = 42
test_data_args: Dict[str, Any]
train_data_args: Dict[str, Any]
trainer_args: Dict[str, Any]
trainer_cls: Any
class ssl_tools.benchmarks.main_mix_style.ResNet1DBase(resnet_block_cls=ResNetBlock, activation_cls=torch.nn.ReLU, input_shape=(6, 60), num_classes=6, num_residual_blocks=5, reduction_ratio=2, learning_rate=0.001)

Bases: SimpleClassificationNet2

Parameters:

  • resnet_block_cls (type)

  • activation_cls (type)

  • input_shape (Tuple[int, int])

  • num_classes (int)

  • num_residual_blocks (int)

  • learning_rate (float)

_calculate_fc_input_features(backbone, input_shape)

Run a single forward pass with a random input to get the number of features after the convolutional layers.

Parameters

backbonetorch.nn.Module

The backbone of the network

input_shapeTuple[int, int, int]

The input shape of the network.

Returns

int

The number of features after the convolutional layers.

Parameters:

  • backbone (torch.nn.Module)

  • input_shape (Tuple[int, int, int])

Return type:

int

class ssl_tools.benchmarks.main_mix_style.ResNet1D_8(*args, **kwargs)

Bases: ResNet1DBase

class ssl_tools.benchmarks.main_mix_style.ResNetBlock(in_channels=64, activation_cls=torch.nn.ReLU, mix_style_factor=False)

Bases: torch.nn.Module

Parameters:

  • in_channels (int)

  • activation_cls (torch.nn.Module)

forward(x)
class ssl_tools.benchmarks.main_mix_style.ResNetSE1D_5(*args, **kwargs)

Bases: ResNet1DBase

class ssl_tools.benchmarks.main_mix_style.ResNetSE1D_8(*args, **kwargs)

Bases: ResNet1DBase

class ssl_tools.benchmarks.main_mix_style.ResNetSEBlock(*args, **kwargs)

Bases: ResNetBlock

class ssl_tools.benchmarks.main_mix_style.SimpleClassificationNet2(backbone, fc, learning_rate=0.001, flatten=True, loss_fn=None, train_metrics=None, val_metrics=None, test_metrics=None)

Bases: ssl_tools.models.nets.simple.SimpleClassificationNet

Parameters:

  • backbone (torch.nn.Module)

  • fc (torch.nn.Module)

  • learning_rate (float)

  • flatten (bool)

  • loss_fn (torch.nn.Module)

  • train_metrics (Dict[str, torch.Tensor])

  • val_metrics (Dict[str, torch.Tensor])

  • test_metrics (Dict[str, torch.Tensor])

single_step(batch, batch_idx, step_name)
Parameters:

  • batch (torch.Tensor)

  • batch_idx (int)

  • step_name (str)

class ssl_tools.benchmarks.main_mix_style.SqueezeAndExcitation1D(in_channels, reduction_ratio=2)

Bases: torch.nn.Module

Parameters:

  • in_channels (int)

  • reduction_ratio (int)

forward(input_tensor)
class ssl_tools.benchmarks.main_mix_style._ResNet1D(input_shape, residual_block_cls=ResNetBlock, activation_cls=torch.nn.ReLU, num_residual_blocks=5, reduction_ratio=2)

Bases: torch.nn.Module

Parameters:

  • input_shape (Tuple[int, int])

  • activation_cls (torch.nn.Module)

  • num_residual_blocks (int)

forward(x)
ssl_tools.benchmarks.main_mix_style._run_experiment_wrapper(experiment_args)
Parameters:

experiment_args (ExperimentArgs)

ssl_tools.benchmarks.main_mix_style.cli_main(experiment)
Parameters:

experiment (ExperimentArgs)

ssl_tools.benchmarks.main_mix_style.conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1)

3x3 convolution with padding

Parameters:

  • in_planes (int)

  • out_planes (int)

  • stride (int)

  • groups (int)

  • dilation (int)

Return type:

torch.nn.Conv2d

ssl_tools.benchmarks.main_mix_style.conv3x3_dynamic(in_planes, out_planes, stride=1, attention_in_channels=None)

3x3 convolution with padding

Parameters:

  • in_planes (int)

  • out_planes (int)

  • stride (int)

  • attention_in_channels (int)

Return type:

dassl.modeling.ops.Conv2dDynamic

ssl_tools.benchmarks.main_mix_style.main_loo()
ssl_tools.benchmarks.main_mix_style.pretty_print_experiment_args(args, indent=4)
Parameters:
Return type:

str

ssl_tools.benchmarks.main_mix_style.run_serial(experiments)
Parameters:

experiments (List[ExperimentArgs])

ssl_tools.benchmarks.main_mix_style.run_using_ray(experiments, ray_address=None)
Parameters: