Source code for monai.networks.blocks.mlp

# Copyright (c) MONAI Consortium
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.

from typing import Tuple, Union

import torch.nn as nn

from monai.networks.layers import get_act_layer
from monai.utils import look_up_option

SUPPORTED_DROPOUT_MODE = {"vit", "swin"}

[docs]class MLPBlock(nn.Module): """ A multi-layer perceptron block, based on: "Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <>" """
[docs] def __init__( self, hidden_size: int, mlp_dim: int, dropout_rate: float = 0.0, act: Union[Tuple, str] = "GELU", dropout_mode="vit", ) -> None: """ Args: hidden_size: dimension of hidden layer. mlp_dim: dimension of feedforward layer. If 0, `hidden_size` will be used. dropout_rate: faction of the input units to drop. act: activation type and arguments. Defaults to GELU. dropout_mode: dropout mode, can be "vit" or "swin". "vit" mode uses two dropout instances as implemented in "swin" corresponds to one instance as implemented in """ super().__init__() if not (0 <= dropout_rate <= 1): raise ValueError("dropout_rate should be between 0 and 1.") mlp_dim = mlp_dim or hidden_size self.linear1 = nn.Linear(hidden_size, mlp_dim) self.linear2 = nn.Linear(mlp_dim, hidden_size) self.fn = get_act_layer(act) self.drop1 = nn.Dropout(dropout_rate) dropout_opt = look_up_option(dropout_mode, SUPPORTED_DROPOUT_MODE) if dropout_opt == "vit": self.drop2 = nn.Dropout(dropout_rate) elif dropout_opt == "swin": self.drop2 = self.drop1 else: raise ValueError(f"dropout_mode should be one of {SUPPORTED_DROPOUT_MODE}")
[docs] def forward(self, x): x = self.fn(self.linear1(x)) x = self.drop1(x) x = self.linear2(x) x = self.drop2(x) return x