core.models.uma.escn_md_block

Copyright (c) Meta Platforms, Inc. and affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Classes

Edgewise	Base class for all neural network modules.
SpectralAtomwise	Base class for all neural network modules.
GridAtomwise	Base class for all neural network modules.
eSCNMD_Block	Base class for all neural network modules.

Module Contents

class core.models.uma.escn_md_block.Edgewise(sphere_channels: int, hidden_channels: int, lmax: int, mmax: int, edge_channels_list, mappingReduced, SO3_grid, cutoff, activation_checkpoint_chunk_size: int | None, act_type: str = 'gate')

Bases: torch.nn.Module

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes:

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call to(), etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

Variables:

training (bool) – Boolean represents whether this module is in training or evaluation mode.

sphere_channels

hidden_channels

lmax

mmax

activation_checkpoint_chunk_size

mappingReduced

SO3_grid

edge_channels_list

act_type

so2_conv_1

so2_conv_2

cutoff

envelope

out_mask

forward(x, x_edge, edge_distance, edge_index, wigner_and_M_mapping, wigner_and_M_mapping_inv, node_offset: int = 0)

forward_chunk(x, x_edge, edge_distance, edge_index, wigner_and_M_mapping, wigner_and_M_mapping_inv, node_offset: int = 0)

class core.models.uma.escn_md_block.SpectralAtomwise(sphere_channels: int, hidden_channels: int, lmax: int, mmax: int, SO3_grid)

Bases: torch.nn.Module

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes:

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call to(), etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

Variables:

training (bool) – Boolean represents whether this module is in training or evaluation mode.

sphere_channels

hidden_channels

lmax

mmax

SO3_grid

scalar_mlp

so3_linear_1

act

so3_linear_2

forward(x)

class core.models.uma.escn_md_block.GridAtomwise(sphere_channels: int, hidden_channels: int, lmax: int, mmax: int, SO3_grid)

Bases: torch.nn.Module

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes:

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call to(), etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

Variables:

training (bool) – Boolean represents whether this module is in training or evaluation mode.

sphere_channels

hidden_channels

lmax

mmax

SO3_grid

grid_mlp

forward(x)

class core.models.uma.escn_md_block.eSCNMD_Block(sphere_channels: int, hidden_channels: int, lmax: int, mmax: int, mappingReduced, SO3_grid, edge_channels_list: list[int], cutoff: float, norm_type: str, act_type: str, ff_type: str, activation_checkpoint_chunk_size: int | None)

Bases: torch.nn.Module

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes:

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call to(), etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

Variables:

training (bool) – Boolean represents whether this module is in training or evaluation mode.

sphere_channels

hidden_channels

lmax

mmax

norm_1

edge_wise

norm_2

forward(x, x_edge, edge_distance, edge_index, wigner_and_M_mapping, wigner_and_M_mapping_inv, sys_node_embedding=None, node_offset: int = 0)