decode.neuralfitter.models package

Submodules

decode.neuralfitter.models.model_param module

class decode.neuralfitter.models.model_param.DoubleMUnet(ch_in, ch_out, ext_features=0, depth_shared=3, depth_union=3, initial_features=64, inter_features=64, activation=ReLU(), use_last_nl=True, norm=None, norm_groups=None, norm_head=None, norm_head_groups=None, pool_mode='Conv2d', upsample_mode='bilinear', skip_gn_level=None)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

apply_detection_nonlin(x)

Apply detection non-linearity. Useful for non-training situations. When BCEWithLogits loss is used, do not use this

during training (because it’s already included in the loss).

Parameters:

o – model output

Return type:

Tensor

apply_nonlin(o)

Apply non-linearity to all but the detection channel.

Parameters:

o (Tensor) –

Return type:

Tensor

forward(x, force_no_p_nl=False)

Parameters:

• x –

• force_no_p_nl –

Returns:

classmethod parse(param, **kwargs)

rescale_last_layer_grad(loss, optimizer)

Rescales the weight as by the last layer’s gradient

Parameters:

• loss –

• optimizer –

Returns:

weight, channelwise loss, channelwise weighted loss

training: bool

class decode.neuralfitter.models.model_param.MLTHeads(in_channels, out_channels, last_kernel, norm, norm_groups, padding, activation)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.model_param.SimpleSMLMNet(ch_in, ch_out, depth=3, initial_features=64, inter_features=64, p_dropout=0.0, activation=ReLU(), use_last_nl=True, norm=None, norm_groups=None, norm_head=None, norm_head_groups=None, pool_mode='StrideConv', upsample_mode='bilinear', skip_gn_level=None)

Bases: UNet2d

Initializes internal Module state, shared by both nn.Module and ScriptModule.

apply_nonlin(o)

Apply non linearity in all the other channels :type o: :param o: :return:

apply_pnl(o)

Apply nonlinearity (sigmoid) to p channel. This is combined during training in the loss function. Only use when not training :type o: :param o: :return:

static check_target(y_tar)

forward(x, force_no_p_nl=False)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

static parse(param)

rescale_last_layer_grad(loss, optimizer)

Parameters:

• loss – non-reduced loss of size N x C x H x W

• optimizer –

Returns:

weight, channelwise loss, channelwise weighted loss

training: bool

decode.neuralfitter.models.model_speced_impl module

class decode.neuralfitter.models.model_speced_impl.SigmaMUNet(ch_in, *, depth_shared, depth_union, initial_features, inter_features, norm=None, norm_groups=None, norm_head=None, norm_head_groups=None, pool_mode='StrideConv', upsample_mode='bilinear', skip_gn_level=None, activation=ReLU(), kaiming_normal=True)

Bases: DoubleMUnet

Initializes internal Module state, shared by both nn.Module and ScriptModule.

apply_detection_nonlin(x)

Apply detection non-linearity. Useful for non-training situations. When BCEWithLogits loss is used, do not use this

during training (because it’s already included in the loss).

Parameters:

o – model output

Return type:

Tensor

apply_nonlin(o)

Apply non-linearity to all but the detection channel.

Parameters:

o (Tensor) –

Return type:

Tensor

bg_ch_ix = [10]

ch_out = 10

forward(x)

Parameters:

• x (Tensor) –

• force_no_p_nl –

Return type:

Tensor

Returns:

mt_heads

Register sigma as parameter such that it is stored in the models state dict and loaded correctly.

out_channels_heads = (1, 4, 4, 1)

p_ch_ix = [0]

classmethod parse(param, **kwargs)

pxyz_mu_ch_ix = slice(1, 5, None)

pxyz_sig_ch_ix = slice(5, 9, None)

sigma_eps_default = 0.001

sigmoid_ch_ix = [0, 1, 5, 6, 7, 8, 9]

tanh_ch_ix = [2, 3, 4]

training: bool

static weight_init(m)

Apply Kaiming normal init. Call this recursively by model.apply(model.weight_init)

Parameters:

m – model

decode.neuralfitter.models.unet_param module

class decode.neuralfitter.models.unet_param.UNet2d(in_channels, out_channels, depth=4, initial_features=64, gain=2, pad_convs=False, norm=None, norm_groups=None, p_dropout=None, final_activation=None, activation=ReLU(), pool_mode='MaxPool', skip_gn_level=None, upsample_mode='bilinear')

Bases: UNetBase

Initializes internal Module state, shared by both nn.Module and ScriptModule.

training: bool

class decode.neuralfitter.models.unet_param.UNetBase(in_channels, out_channels, depth=4, initial_features=64, gain=2, pad_convs=False, norm=None, norm_groups=None, p_dropout=None, final_activation=None, activation=ReLU(), pool_mode='MaxPool', skip_gn_level=None, upsample_mode='bilinear')

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(input)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

forward_parts(parts)

norms = ('BatchNorm', 'GroupNorm')

pool_modules = ('MaxPool', 'StrideConv')

training: bool

class decode.neuralfitter.models.unet_param.Upsample(scale_factor, mode='nearest', in_channels=None, out_channels=None, align_corners=False, ndim=3)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(input)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

decode.neuralfitter.models.unet_param.get_activation(activation)

Get activation from str or nn.Module

decode.neuralfitter.models.unet_parts module

class decode.neuralfitter.models.unet_parts.Upsample(scale_factor, mode, align_corners)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.unet_parts.double_conv(in_ch, out_ch)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.unet_parts.down(in_ch, out_ch)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.unet_parts.down_3d(in_ch, out_ch)

Bases: down

Initializes internal Module state, shared by both nn.Module and ScriptModule.

training: bool

class decode.neuralfitter.models.unet_parts.inconv(in_ch, out_ch)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.unet_parts.inconv_3d(in_ch, out_ch)

Bases: inconv

Initializes internal Module state, shared by both nn.Module and ScriptModule.

training: bool

class decode.neuralfitter.models.unet_parts.outconv(in_ch, out_ch)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class decode.neuralfitter.models.unet_parts.up(in_ch, out_ch, bilinear=True)

Bases: Module

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(x1, x2)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

Module contents