# Copyright 2020 - 2021 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
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Tuple
import torch.nn as nn
from monai.networks.blocks.patchembedding import PatchEmbeddingBlock
from monai.networks.blocks.transformerblock import TransformerBlock
[docs]class ViT(nn.Module):
"""
Vision Transformer (ViT), based on: "Dosovitskiy et al.,
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <https://arxiv.org/abs/2010.11929>"
"""
def __init__(
self,
in_channels: int,
img_size: Tuple[int, int, int],
patch_size: Tuple[int, int, int],
hidden_size: int,
mlp_dim: int,
num_layers: int,
num_heads: int,
pos_embed: str,
classification: bool,
num_classes: int = 2,
dropout_rate: float = 0.0,
) -> None:
"""
Args:
in_channels: dimension of input channels.
img_size: dimension of input image.
patch_size: dimension of patch size.
hidden_size: dimension of hidden layer.
mlp_dim: dimension of feedforward layer.
num_layers: number of transformer blocks.
num_heads: number of attention heads.
pos_embed: position embedding layer type.
classification: bool argument to determine if classification is used.
num_classes: number of classes if classification is used.
dropout_rate: faction of the input units to drop.
"""
super().__init__()
if not (0 <= dropout_rate <= 1):
raise AssertionError("dropout_rate should be between 0 and 1.")
if hidden_size % num_heads != 0:
raise AssertionError("hidden size should be divisible by num_heads.")
if pos_embed not in ["conv", "perceptron"]:
raise KeyError(f"Position embedding layer of type {pos_embed} is not supported.")
self.classification = classification
self.patch_embedding = PatchEmbeddingBlock(
in_channels, img_size, patch_size, hidden_size, num_heads, pos_embed, dropout_rate
)
self.blocks = nn.ModuleList(
[TransformerBlock(hidden_size, mlp_dim, num_heads, dropout_rate) for i in range(num_layers)]
)
self.norm = nn.LayerNorm(hidden_size)
if self.classification:
self.classification_head = nn.Linear(hidden_size, num_classes)
[docs] def forward(self, x):
x = self.patch_embedding(x)
hidden_states_out = []
for blk in self.blocks:
x = blk(x)
hidden_states_out.append(x)
x = self.norm(x)
if self.classification:
x = self.classification_head(x[:, 0])
return x, hidden_states_out