Inheritance diagram for Model:

[legend]

Collaboration diagram for Model:

[legend]

Public Member Functions
	__init__ (self)
	Constructor.

	init_weights (self)
	Initialize.

	forward (self, z, y)
	forward

Public Attributes
	ch = G_ch
	Channel width multiplier.

	G_depth = G_depth
	Number of resblocks per stage.

	dim_z = dim_z
	Dimensionality of the latent space.

	bottom_width = bottom_width
	The initial spatial dimensions.

	H_base = H_base
	The initial harizontal dimension.

	resolution = resolution
	Resolution of the output.

	kernel_size = G_kernel_size
	Kernel size?

	attention = G_attn
	Attention?

	n_classes = n_classes
	number of classes, for use in categorical conditional generation

	G_shared = G_shared
	Use shared embeddings?

int	shared_dim = shared_dim if shared_dim > 0 else dim_z
	Dimensionality of the shared embedding?

# y and z are[bs, 128] dimensional	hier = hier
	Hierarchical latent space?

	cross_replica = cross_replica
	Cross replica batchnorm?

	mybn = mybn
	Use my batchnorm?

	activation = torch.nn.ReLU(inplace=True)
	activation

str	init = G_init
	Initialization style.

str	G_param = G_param
	Parameterization style.

	norm_style = norm_style
	Normalization style.

	BN_eps = BN_eps
	Epsilon for BatchNorm?

	SN_eps = SN_eps
	Epsilon for Spectral Norm?

	fp16 = G_fp16
	fp16?

	arch = G_arch(self.ch, self.attention)[resolution]
	Architecture dict.

	RRM_prx_G = RRM_prx_G
	RRM_prx_G.

	n_head_G = n_head_G
	n_head_G

	which_conv
	which conv

	which_linear
	which linear

	which_embedding = nn.Embedding
	which embedding

	which_bn
	which bn

tuple	shared
	shared

	RR_G
	RRM on proxy embeddings.

	linear
	First linear layer.

list	blocks = []
	blocks

	output_layer
	output layer

	lr = G_lr
	lr

	B1 = G_B1
	B1.

	B2 = G_B2
	B2.

	adam_eps = adam_eps
	adam_eps

	optim
	optim

	lr_sched = None
	lr sched

int	param_count = 0
	parameter count

Detailed Description

Generator subclass
default initializing with CONFIG dict

Definition at line 1459 of file ieagan.py.

Constructor & Destructor Documentation

◆ init()

__init__ ( self )

Constructor.

Definition at line 1466 of file ieagan.py.

    def __init__(self):
        super().__init__(**CONFIG)

Member Function Documentation

◆ forward()

forward	(	self,
		z,
		y )

inherited

forward

Definition at line 1435 of file ieagan.py.

    def forward(self, z, y):
        y = self.shared(y)
        # If relational embedding
        if self.RRM_prx_G:
            y = self.RR_G(y.unsqueeze(0)).squeeze(0)
            # y = F.normalize(y, dim=1)
        # If hierarchical, concatenate zs and ys
        if self.hier:  # y and z are [bs,128] dimensional
            z = torch.cat([y, z], 1)
            y = z
        # First linear layer
        h = self.linear(z)  # ([bs,256]-->[bs,24576])
        # Reshape
        h = h.view(h.size(0), -1, self.bottom_width, self.bottom_width * self.H_base)
        # Loop over blocks
        for _, blocklist in enumerate(self.blocks):
            # Second inner loop in case block has multiple layers
            for block in blocklist:
                h = block(h, y)
 
        # Apply batchnorm-relu-conv-tanh at output
        return torch.tanh(self.output_layer(h))
 
 

◆ init_weights()

init_weights ( self )

inherited

Initialize.

Definition at line 1412 of file ieagan.py.

    def init_weights(self):
        
        self.param_count = 0
        for module in self.modules():
            if (
                isinstance(module, nn.Conv2d)
                or isinstance(module, nn.Linear)
                or isinstance(module, nn.Embedding)
            ):
                if self.init == "ortho":
                    init.orthogonal_(module.weight)
                elif self.init == "N02":
                    init.normal_(module.weight, 0, 0.02)
                elif self.init in ["glorot", "xavier"]:
                    init.xavier_uniform_(module.weight)
                else:
                    print("Init style not recognized...")
                self.param_count += sum(
                    [p.data.nelement() for p in module.parameters()]
                )
        print(f"Param count for G's initialized parameters: {self.param_count}")
 

Member Data Documentation

◆ activation

activation = torch.nn.ReLU(inplace=True)

inherited

activation

Definition at line 1201 of file ieagan.py.

◆ adam_eps

adam_eps = adam_eps

inherited

adam_eps

Definition at line 1372 of file ieagan.py.

◆ arch

arch = G_arch(self.ch, self.attention)[resolution]

inherited

Architecture dict.

Definition at line 1221 of file ieagan.py.

◆ attention

attention = G_attn

inherited

Attention?

Definition at line 1185 of file ieagan.py.

◆ B1

B1 = G_B1

inherited

B1.

Definition at line 1368 of file ieagan.py.

◆ B2

B2 = G_B2

inherited

B2.

Definition at line 1370 of file ieagan.py.

◆ blocks

blocks = []

inherited

blocks

Definition at line 1299 of file ieagan.py.

◆ BN_eps

BN_eps = BN_eps

inherited

Epsilon for BatchNorm?

Definition at line 1215 of file ieagan.py.

◆ bottom_width

bottom_width = bottom_width

inherited

The initial spatial dimensions.

Definition at line 1177 of file ieagan.py.

◆ ch

ch = G_ch

inherited

Channel width multiplier.

Definition at line 1171 of file ieagan.py.

◆ cross_replica

cross_replica = cross_replica

inherited

Cross replica batchnorm?

Definition at line 1195 of file ieagan.py.

◆ dim_z

dim_z = dim_z

inherited

Dimensionality of the latent space.

Definition at line 1175 of file ieagan.py.

◆ fp16

fp16 = G_fp16

inherited

fp16?

Definition at line 1219 of file ieagan.py.

◆ G_depth

G_depth = G_depth

inherited

Number of resblocks per stage.

Definition at line 1173 of file ieagan.py.

◆ G_param

str G_param = G_param

inherited

Parameterization style.

Definition at line 1211 of file ieagan.py.

◆ G_shared

G_shared = G_shared

inherited

Use shared embeddings?

Definition at line 1189 of file ieagan.py.

◆ H_base

H_base = H_base

inherited

The initial harizontal dimension.

Definition at line 1179 of file ieagan.py.

◆ hier

# y and z are [bs,128] dimensional hier = hier

inherited

Hierarchical latent space?

Definition at line 1193 of file ieagan.py.

◆ init

str init = G_init

inherited

Initialization style.

Definition at line 1209 of file ieagan.py.

◆ kernel_size

kernel_size = G_kernel_size

inherited

Kernel size?

Definition at line 1183 of file ieagan.py.

◆ linear

linear

inherited

Initial value:

=  self.which_linear(
            self.dim_z + self.shared_dim,
            self.arch["in_channels"][0] * ((self.bottom_width**2) * self.H_base),
        )

First linear layer.

Definition at line 1290 of file ieagan.py.

◆ lr

lr = G_lr

inherited

lr

Definition at line 1366 of file ieagan.py.

◆ lr_sched

lr_sched = None

inherited

lr sched

Definition at line 1396 of file ieagan.py.

◆ mybn

mybn = mybn

inherited

Use my batchnorm?

Definition at line 1197 of file ieagan.py.

◆ n_classes

n_classes = n_classes

inherited

number of classes, for use in categorical conditional generation

Definition at line 1187 of file ieagan.py.

◆ n_head_G

n_head_G = n_head_G

inherited

n_head_G

Definition at line 1225 of file ieagan.py.

◆ norm_style

norm_style = norm_style

inherited

Normalization style.

Definition at line 1213 of file ieagan.py.

◆ optim

optim

inherited

Initial value:

=  utils.Adam16(
                params=self.parameters(),
                lr=self.lr,
                betas=(self.B1, self.B2),
                weight_decay=0,
                eps=self.adam_eps,
            )

optim

Definition at line 1377 of file ieagan.py.

◆ output_layer

output_layer

inherited

Initial value:

=  nn.Sequential(
            bn(
                self.arch["out_channels"][-1],
                cross_replica=self.cross_replica,
                mybn=self.mybn,
            ),
            self.activation,
            self.which_conv(self.arch["out_channels"][-1], 1),
        )

output layer

Definition at line 1347 of file ieagan.py.

◆ param_count

int param_count = 0

inherited

parameter count

Definition at line 1414 of file ieagan.py.

◆ resolution

resolution = resolution

inherited

Resolution of the output.

Definition at line 1181 of file ieagan.py.

◆ RR_G

RR_G

inherited

Initial value:

=  RelationalReasoning(
                num_layers=1,
                input_dim=128,
                dim_feedforward=128,
                which_linear=nn.Linear,
                num_heads=self.n_head_G,
                dropout=0.0,
                hidden_dim=128,
            )

RRM on proxy embeddings.

Definition at line 1279 of file ieagan.py.

◆ RRM_prx_G

RRM_prx_G = RRM_prx_G

inherited

RRM_prx_G.

Definition at line 1223 of file ieagan.py.

◆ shared

tuple shared

inherited

Initial value:

=  (
            self.which_embedding(n_classes, self.shared_dim)
            if G_shared
            else identity()
        )

shared

Definition at line 1271 of file ieagan.py.

◆ shared_dim

int shared_dim = shared_dim if shared_dim > 0 else dim_z

inherited

Dimensionality of the shared embedding?

Unused if not using G_shared

Definition at line 1191 of file ieagan.py.

◆ SN_eps

SN_eps = SN_eps

inherited

Epsilon for Spectral Norm?

Definition at line 1217 of file ieagan.py.

◆ which_bn

which_bn

inherited

Initial value:

=  functools.partial(
            ccbn,
            which_linear=bn_linear,
            cross_replica=self.cross_replica,
            mybn=self.mybn,
            input_size=(
                self.shared_dim + self.dim_z if self.G_shared else self.n_classes
            ),
            norm_style=self.norm_style,
            eps=self.BN_eps,
        )

which bn

Definition at line 1259 of file ieagan.py.

◆ which_conv

which_conv

inherited

Initial value:

=  functools.partial(
                SNConv2d,
                kernel_size=3,
                padding=1,
                num_svs=num_G_SVs,
                num_itrs=num_G_SV_itrs,
                eps=self.SN_eps,
            )

which conv

Definition at line 1230 of file ieagan.py.

◆ which_embedding

which_embedding = nn.Embedding

inherited

which embedding

Definition at line 1252 of file ieagan.py.

◆ which_linear

which_linear

inherited

Initial value:

=  functools.partial(
                SNLinear,
                num_svs=num_G_SVs,
                num_itrs=num_G_SV_itrs,
                eps=self.SN_eps,
            )

which linear

Definition at line 1239 of file ieagan.py.

The documentation for this class was generated from the following file:

pxd/scripts/pxd/background_generator/models/ieagan.py

Public Member Functions

Public Attributes

Detailed Description

Constructor & Destructor Documentation

◆ __init__()

Member Function Documentation

◆ forward()

◆ init_weights()

Member Data Documentation

◆ activation

◆ adam_eps

◆ arch

◆ attention

◆ B1

◆ B2

◆ blocks

◆ BN_eps

◆ bottom_width

◆ ch

◆ cross_replica

◆ dim_z

◆ fp16

◆ G_depth

◆ G_param

◆ G_shared

◆ H_base

◆ hier

◆ init

◆ kernel_size

◆ linear

◆ lr

◆ lr_sched

◆ mybn

◆ n_classes

◆ n_head_G

◆ norm_style

◆ optim

◆ output_layer

◆ param_count

◆ resolution

◆ RR_G

◆ RRM_prx_G

◆ shared

◆ shared_dim

◆ SN_eps

◆ which_bn

◆ which_conv

◆ which_embedding

◆ which_linear

◆ init()