intent to build (#210)

complete SepViT, from bytedance AI labs

README.md

@@ -18,13 +18,18 @@
 - [Twins SVT](#twins-svt)
 - [CrossFormer](#crossformer)
 - [RegionViT](#regionvit)
+- [ScalableViT](#scalablevit)
+- [SepViT](#sepvit)
 - [NesT](#nest)
 - [MobileViT](#mobilevit)
 - [Masked Autoencoder](#masked-autoencoder)
 - [Simple Masked Image Modeling](#simple-masked-image-modeling)
 - [Masked Patch Prediction](#masked-patch-prediction)
 - [Adaptive Token Sampling](#adaptive-token-sampling)
+- [Patch Merger](#patch-merger)
 - [Vision Transformer for Small Datasets](#vision-transformer-for-small-datasets)
+- [Parallel ViT](#parallel-vit)
+- [Learnable Memory ViT](#learnable-memory-vit)
 - [Dino](#dino)
 - [Accessing Attention](#accessing-attention)
 - [Research Ideas](#research-ideas)
@@ -42,6 +47,8 @@ For a Pytorch implementation with pretrained models, please see Ross Wightman's
 
 The official Jax repository is <a href="https://github.com/google-research/vision_transformer">here</a>.
 
+A tensorflow2 translation also exists <a href="https://github.com/taki0112/vit-tensorflow">here</a>, created by research scientist <a href="https://github.com/taki0112">Junho Kim</a>! 🙏
+
 ## Install
 
 ```bash
@@ -238,6 +245,7 @@ preds = v(img) # (1, 1000)
 ```
 
 ## CCT
+
 <img src="https://raw.githubusercontent.com/SHI-Labs/Compact-Transformers/main/images/model_sym.png" width="400px"></img>
 
 <a href="https://arxiv.org/abs/2104.05704">CCT</a> proposes compact transformers
@@ -249,22 +257,25 @@ You can use this with two methods
 import torch
 from vit_pytorch.cct import CCT
 
-model = CCT(
-        img_size=224,
-        embedding_dim=384,
-        n_conv_layers=2,
-        kernel_size=7,
-        stride=2,
-        padding=3,
-        pooling_kernel_size=3,
-        pooling_stride=2,
-        pooling_padding=1,
-        num_layers=14,
-        num_heads=6,
-        mlp_radio=3.,
-        num_classes=1000,
-        positional_embedding='learnable', # ['sine', 'learnable', 'none']
-        )
+cct = CCT(
+    img_size = (224, 448),
+    embedding_dim = 384,
+    n_conv_layers = 2,
+    kernel_size = 7,
+    stride = 2,
+    padding = 3,
+    pooling_kernel_size = 3,
+    pooling_stride = 2,
+    pooling_padding = 1,
+    num_layers = 14,
+    num_heads = 6,
+    mlp_radio = 3.,
+    num_classes = 1000,
+    positional_embedding = 'learnable', # ['sine', 'learnable', 'none']
+)
+
+img = torch.randn(1, 3, 224, 448)
+pred = cct(img) # (1, 1000)
 ```
 
 Alternatively you can use one of several pre-defined models `[2,4,6,7,8,14,16]`
@@ -275,23 +286,23 @@ and the embedding dimension.
 import torch
 from vit_pytorch.cct import cct_14
 
-model = cct_14(
-        img_size=224,
-        n_conv_layers=1,
-        kernel_size=7,
-        stride=2,
-        padding=3,
-        pooling_kernel_size=3,
-        pooling_stride=2,
-        pooling_padding=1,
-        num_classes=1000,
-        positional_embedding='learnable', # ['sine', 'learnable', 'none']  
-        )
+cct = cct_14(
+    img_size = 224,
+    n_conv_layers = 1,
+    kernel_size = 7,
+    stride = 2,
+    padding = 3,
+    pooling_kernel_size = 3,
+    pooling_stride = 2,
+    pooling_padding = 1,
+    num_classes = 1000,
+    positional_embedding = 'learnable', # ['sine', 'learnable', 'none']
+)
 ```
+
 <a href="https://github.com/SHI-Labs/Compact-Transformers">Official
 Repository</a> includes links to pretrained model checkpoints.
 
-
 ## Cross ViT
 
 <img src="./images/cross_vit.png" width="400px"></img>
@@ -524,6 +535,67 @@ img = torch.randn(1, 3, 224, 224)
 pred = model(img) # (1, 1000)
 ```
 
+## ScalableViT
+
+<img src="./images/scalable-vit-1.png" width="400px"></img>
+
+<img src="./images/scalable-vit-2.png" width="400px"></img>
+
+This Bytedance AI <a href="https://arxiv.org/abs/2203.10790">paper</a> proposes the Scalable Self Attention (SSA) and the Interactive Windowed Self Attention (IWSA) modules. The SSA alleviates the computation needed at earlier stages by reducing the key / value feature map by some factor (`reduction_factor`), while modulating the dimension of the queries and keys (`ssa_dim_key`). The IWSA performs self attention within local windows, similar to other vision transformer papers. However, they add a residual of the values, passed through a convolution of kernel size 3, which they named Local Interactive Module (LIM).
+
+They make the claim in this paper that this scheme outperforms Swin Transformer, and also demonstrate competitive performance against Crossformer.
+
+You can use it as follows (ex. ScalableViT-S)
+
+```python
+import torch
+from vit_pytorch.scalable_vit import ScalableViT
+
+model = ScalableViT(
+    num_classes = 1000,
+    dim = 64,                               # starting model dimension. at every stage, dimension is doubled
+    heads = (2, 4, 8, 16),                  # number of attention heads at each stage
+    depth = (2, 2, 20, 2),                  # number of transformer blocks at each stage
+    ssa_dim_key = (40, 40, 40, 32),         # the dimension of the attention keys (and queries) for SSA. in the paper, they represented this as a scale factor on the base dimension per key (ssa_dim_key / dim_key)
+    reduction_factor = (8, 4, 2, 1),        # downsampling of the key / values in SSA. in the paper, this was represented as (reduction_factor ** -2)
+    window_size = (64, 32, None, None),     # window size of the IWSA at each stage. None means no windowing needed
+    dropout = 0.1,                          # attention and feedforward dropout
+)
+
+img = torch.randn(1, 3, 256, 256)
+
+preds = model(img) # (1, 1000)
+```
+
+## SepViT
+
+<img src="./images/sep-vit.png" width="400px"></img>
+
+Another <a href="https://arxiv.org/abs/2203.15380">Bytedance AI paper</a>, it proposes a depthwise-pointwise self-attention layer that seems largely inspired by mobilenet's depthwise-separable convolution. The most interesting aspect is the reuse of the feature map from the depthwise self-attention stage as the values for the pointwise self-attention, as shown in the diagram above.
+
+I have decided to include only the version of `SepViT` with this specific self-attention layer, as the grouped attention layers are not remarkable nor novel, and the authors were not clear on how they treated the window tokens for the group self-attention layer. Besides, it seems like with `DSSA` layer alone, they were able to beat Swin.
+
+ex. SepViT-Lite
+
+```python
+import torch
+from vit_pytorch.sep_vit import SepViT
+
+v = SepViT(
+    num_classes = 1000,
+    dim = 32,               # dimensions of first stage, which doubles every stage (32, 64, 128, 256) for SepViT-Lite
+    dim_head = 32,          # attention head dimension
+    heads = (1, 2, 4, 8),   # number of heads per stage
+    depth = (1, 2, 6, 2),   # number of transformer blocks per stage
+    window_size = 7,        # window size of DSS Attention block
+    dropout = 0.1           # dropout
+)
+
+img = torch.randn(1, 3, 224, 224)
+
+preds = v(img) # (1, 1000)
+```
+
 ## NesT
 
 <img src="./images/nest.png" width="400px"></img>
@@ -732,12 +804,58 @@ v = ViT(
 
 img = torch.randn(4, 3, 256, 256)
 
-preds = v(img) # (1, 1000)
+preds = v(img) # (4, 1000)
 
 # you can also get a list of the final sampled patch ids
 # a value of -1 denotes padding
 
-preds, token_ids = v(img, return_sampled_token_ids = True) # (1, 1000), (1, <=8)
+preds, token_ids = v(img, return_sampled_token_ids = True) # (4, 1000), (4, <=8)
+```
+
+## Patch Merger
+
+
+<img src="./images/patch_merger.png" width="400px"></img>
+
+This <a href="https://arxiv.org/abs/2202.12015">paper</a> proposes a simple module (Patch Merger) for reducing the number of tokens at any layer of a vision transformer without sacrificing performance.
+
+```python
+import torch
+from vit_pytorch.vit_with_patch_merger import ViT
+
+v = ViT(
+    image_size = 256,
+    patch_size = 16,
+    num_classes = 1000,
+    dim = 1024,
+    depth = 12,
+    heads = 8,
+    patch_merge_layer = 6,        # at which transformer layer to do patch merging
+    patch_merge_num_tokens = 8,   # the output number of tokens from the patch merge
+    mlp_dim = 2048,
+    dropout = 0.1,
+    emb_dropout = 0.1
+)
+
+img = torch.randn(4, 3, 256, 256)
+
+preds = v(img) # (4, 1000)
+```
+
+One can also use the `PatchMerger` module by itself
+
+```python
+import torch
+from vit_pytorch.vit_with_patch_merger import PatchMerger
+
+merger = PatchMerger(
+    dim = 1024,
+    num_tokens_out = 8   # output number of tokens
+)
+
+features = torch.randn(4, 256, 1024) # (batch, num tokens, dimension)
+
+out = merger(features) # (4, 8, 1024)
 ```
 
 ## Vision Transformer for Small Datasets
@@ -786,6 +904,92 @@ img = torch.randn(4, 3, 256, 256)
 tokens = spt(img) # (4, 256, 1024)
 ```
 
+## Parallel ViT
+
+<img src="./images/parallel-vit.png" width="350px"></img>
+
+This <a href="https://arxiv.org/abs/2203.09795">paper</a> propose parallelizing multiple attention and feedforward blocks per layer (2 blocks), claiming that it is easier to train without loss of performance.
+
+You can try this variant as follows
+
+```python
+import torch
+from vit_pytorch.parallel_vit import ViT
+
+v = ViT(
+    image_size = 256,
+    patch_size = 16,
+    num_classes = 1000,
+    dim = 1024,
+    depth = 6,
+    heads = 8,
+    mlp_dim = 2048,
+    num_parallel_branches = 2,  # in paper, they claimed 2 was optimal
+    dropout = 0.1,
+    emb_dropout = 0.1
+)
+
+img = torch.randn(4, 3, 256, 256)
+
+preds = v(img) # (4, 1000)
+```
+
+## Learnable Memory ViT
+
+<img src="./images/learnable-memory-vit.png" width="350px"></img>
+
+This <a href="https://arxiv.org/abs/2203.15243">paper</a> shows that adding learnable memory tokens at each layer of a vision transformer can greatly enhance fine-tuning results (in addition to learnable task specific CLS token and adapter head).
+
+You can use this with a specially modified `ViT` as follows
+
+```python
+import torch
+from vit_pytorch.learnable_memory_vit import ViT, Adapter
+
+# normal base ViT
+
+v = ViT(
+    image_size = 256,
+    patch_size = 16,
+    num_classes = 1000,
+    dim = 1024,
+    depth = 6,
+    heads = 8,
+    mlp_dim = 2048,
+    dropout = 0.1,
+    emb_dropout = 0.1
+)
+
+img = torch.randn(4, 3, 256, 256)
+logits = v(img) # (4, 1000)
+
+# do your usual training with ViT
+# ...
+
+
+# then, to finetune, just pass the ViT into the Adapter class
+# you can do this for multiple Adapters, as shown below
+
+adapter1 = Adapter(
+    vit = v,
+    num_classes = 2,               # number of output classes for this specific task
+    num_memories_per_layer = 5     # number of learnable memories per layer, 10 was sufficient in paper
+)
+
+logits1 = adapter1(img) # (4, 2) - predict 2 classes off frozen ViT backbone with learnable memories and task specific head
+
+# yet another task to finetune on, this time with 4 classes
+
+adapter2 = Adapter(
+    vit = v,
+    num_classes = 4,
+    num_memories_per_layer = 10
+)
+
+logits2 = adapter2(img) # (4, 4) - predict 4 classes off frozen ViT backbone with learnable memories and task specific head
+
+```
+
 ## Dino
 
 <img src="./images/dino.png" width="350px"></img>
@@ -1294,6 +1498,52 @@ Coming from computer vision and new to transformers? Here are some resources tha
 }
 ```
 
+```bibtex
+@misc{renggli2022learning,
+    title   = {Learning to Merge Tokens in Vision Transformers},
+    author  = {Cedric Renggli and André Susano Pinto and Neil Houlsby and Basil Mustafa and Joan Puigcerver and Carlos Riquelme},
+    year    = {2022},
+    eprint  = {2202.12015},
+    archivePrefix = {arXiv},
+    primaryClass = {cs.CV}
+}
+```
+
+```bibtex
+@misc{yang2022scalablevit,
+    title   = {ScalableViT: Rethinking the Context-oriented Generalization of Vision Transformer}, 
+    author  = {Rui Yang and Hailong Ma and Jie Wu and Yansong Tang and Xuefeng Xiao and Min Zheng and Xiu Li},
+    year    = {2022},
+    eprint  = {2203.10790},
+    archivePrefix = {arXiv},
+    primaryClass = {cs.CV}
+}
+```
+
+```bibtex
+@inproceedings{Touvron2022ThreeTE,
+    title   = {Three things everyone should know about Vision Transformers},
+    author  = {Hugo Touvron and Matthieu Cord and Alaaeldin El-Nouby and Jakob Verbeek and Herv'e J'egou},
+    year    = {2022}
+}
+```
+
+```bibtex
+@inproceedings{Sandler2022FinetuningIT,
+    title   = {Fine-tuning Image Transformers using Learnable Memory},
+    author  = {Mark Sandler and Andrey Zhmoginov and Max Vladymyrov and Andrew Jackson},
+    year    = {2022}
+}
+```
+
+```bibtex
+@inproceedings{Li2022SepViTSV,
+    title   = {SepViT: Separable Vision Transformer},
+    author  = {Wei Li and Xing Wang and Xin Xia and Jie Wu and Xuefeng Xiao and Minghang Zheng and Shiping Wen},
+    year    = {2022}
+}
+```
+
 ```bibtex
 @misc{vaswani2017attention,
     title   = {Attention Is All You Need},

setup.py

@@ -3,7 +3,7 @@ from setuptools import setup, find_packages
 setup(
   name = 'vit-pytorch',
   packages = find_packages(exclude=['examples']),
-  version = '0.26.6',
+  version = '0.32.0',
   license='MIT',
   description = 'Vision Transformer (ViT) - Pytorch',
   author = 'Phil Wang',
@@ -15,7 +15,7 @@ setup(
     'image recognition'
   ],
   install_requires=[
-    'einops>=0.3',
+    'einops>=0.4.1',
     'torch>=1.6',
     'torchvision'
   ],

vit_pytorch/ats_vit.py

@@ -139,6 +139,8 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
         self.output_num_tokens = output_num_tokens
@@ -163,6 +165,7 @@ class Attention(nn.Module):
             dots = dots.masked_fill(~dots_mask, mask_value)
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         sampled_token_ids = None
 

vit_pytorch/cait.py

@@ -76,6 +76,7 @@ class Attention(nn.Module):
         self.to_kv = nn.Linear(dim, inner_dim * 2, bias = False)
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
 
         self.mix_heads_pre_attn = nn.Parameter(torch.randn(heads, heads))
         self.mix_heads_post_attn = nn.Parameter(torch.randn(heads, heads))
@@ -96,7 +97,10 @@ class Attention(nn.Module):
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
 
         dots = einsum('b h i j, h g -> b g i j', dots, self.mix_heads_pre_attn)    # talking heads, pre-softmax
+
         attn = self.attend(dots)
+        attn = self.dropout(attn)
+
         attn = einsum('b h i j, h g -> b g i j', attn, self.mix_heads_post_attn)   # talking heads, post-softmax
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)

vit_pytorch/cct.py

@@ -2,7 +2,13 @@ import torch
 import torch.nn as nn
 import torch.nn.functional as F
 
-# Pre-defined CCT Models
+# helpers
+
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
+# CCT Models
+
 __all__ = ['cct_2', 'cct_4', 'cct_6', 'cct_7', 'cct_8', 'cct_14', 'cct_16']
 
 
@@ -55,8 +61,8 @@ def _cct(num_layers, num_heads, mlp_ratio, embedding_dim,
                padding=padding,
                *args, **kwargs)
 
+# modules
 
-# Modules
 class Attention(nn.Module):
     def __init__(self, dim, num_heads=8, attention_dropout=0.1, projection_dropout=0.1):
         super().__init__()
@@ -308,6 +314,7 @@ class CCT(nn.Module):
                  pooling_padding=1,
                  *args, **kwargs):
         super(CCT, self).__init__()
+        img_height, img_width = pair(img_size)
 
         self.tokenizer = Tokenizer(n_input_channels=n_input_channels,
                                    n_output_channels=embedding_dim,
@@ -324,8 +331,8 @@ class CCT(nn.Module):
 
         self.classifier = TransformerClassifier(
             sequence_length=self.tokenizer.sequence_length(n_channels=n_input_channels,
-                                                           height=img_size,
-                                                           width=img_size),
+                                                           height=img_height,
+                                                           width=img_width),
             embedding_dim=embedding_dim,
             seq_pool=True,
             dropout_rate=0.,
@@ -336,4 +343,3 @@ class CCT(nn.Module):
     def forward(self, x):
         x = self.tokenizer(x)
         return self.classifier(x)
-

vit_pytorch/cross_vit.py

@@ -48,6 +48,8 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
         self.to_q = nn.Linear(dim, inner_dim, bias = False)
         self.to_kv = nn.Linear(dim, inner_dim * 2, bias = False)
 
@@ -69,6 +71,7 @@ class Attention(nn.Module):
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')

vit_pytorch/crossformer.py

@@ -95,6 +95,9 @@ class Attention(nn.Module):
         self.window_size = window_size
 
         self.norm = LayerNorm(dim)
+
+        self.dropout = nn.Dropout(dropout)
+
         self.to_qkv = nn.Conv2d(dim, inner_dim * 3, 1, bias = False)
         self.to_out = nn.Conv2d(inner_dim, dim, 1)
 
@@ -151,6 +154,7 @@ class Attention(nn.Module):
         # attend
 
         attn = sim.softmax(dim = -1)
+        attn = self.dropout(attn)
 
         # merge heads
 

vit_pytorch/cvt.py

@@ -76,6 +76,7 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
 
         self.to_q = DepthWiseConv2d(dim, inner_dim, proj_kernel, padding = padding, stride = 1, bias = False)
         self.to_kv = DepthWiseConv2d(dim, inner_dim * 2, proj_kernel, padding = padding, stride = kv_proj_stride, bias = False)
@@ -94,6 +95,7 @@ class Attention(nn.Module):
         dots = einsum('b i d, b j d -> b i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b i j, b j d -> b i d', attn, v)
         out = rearrange(out, '(b h) (x y) d -> b (h d) x y', h = h, y = y)

vit_pytorch/deepvit.py

@@ -42,6 +42,8 @@ class Attention(nn.Module):
 
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
+        self.dropout = nn.Dropout(dropout)
+
         self.reattn_weights = nn.Parameter(torch.randn(heads, heads))
 
         self.reattn_norm = nn.Sequential(
@@ -64,6 +66,7 @@ class Attention(nn.Module):
 
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
         attn = dots.softmax(dim=-1)
+        attn = self.dropout(attn)
 
         # re-attention
 

vit_pytorch/efficient.py

@@ -3,12 +3,16 @@ from torch import nn
 from einops import rearrange, repeat
 from einops.layers.torch import Rearrange
 
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
 class ViT(nn.Module):
     def __init__(self, *, image_size, patch_size, num_classes, dim, transformer, pool = 'cls', channels = 3):
         super().__init__()
-        assert image_size % patch_size == 0, 'image dimensions must be divisible by the patch size'
+        image_size_h, image_size_w = pair(image_size)
+        assert image_size_h % patch_size == 0 and image_size_w % patch_size == 0, 'image dimensions must be divisible by the patch size'
         assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)'
-        num_patches = (image_size // patch_size) ** 2
+        num_patches = (image_size_h // patch_size) * (image_size_w // patch_size)
         patch_dim = channels * patch_size ** 2
 
         self.to_patch_embedding = nn.Sequential(

vit_pytorch/learnable_memory_vit.py

@@ -0,0 +1,216 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from einops import rearrange, repeat
+from einops.layers.torch import Rearrange
+
+# helpers
+
+def exists(val):
+    return val is not None
+
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
+# controlling freezing of layers
+
+def set_module_requires_grad_(module, requires_grad):
+    for param in module.parameters():
+        param.requires_grad = requires_grad
+
+def freeze_all_layers_(module):
+    set_module_requires_grad_(module, False)
+
+def unfreeze_all_layers_(module):
+    set_module_requires_grad_(module, True)
+
+# classes
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, hidden_dim, dropout = 0.):
+        super().__init__()
+        self.net = nn.Sequential(
+            nn.LayerNorm(dim),
+            nn.Linear(dim, hidden_dim),
+            nn.GELU(),
+            nn.Dropout(dropout),
+            nn.Linear(hidden_dim, dim),
+            nn.Dropout(dropout)
+        )
+    def forward(self, x):
+        return self.net(x)
+
+class Attention(nn.Module):
+    def __init__(self, dim, heads = 8, dim_head = 64, dropout = 0.):
+        super().__init__()
+        inner_dim = dim_head *  heads
+
+        self.heads = heads
+        self.scale = dim_head ** -0.5
+        self.norm = nn.LayerNorm(dim)
+
+        self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
+        self.to_q = nn.Linear(dim, inner_dim, bias = False)
+        self.to_kv = nn.Linear(dim, inner_dim * 2, bias = False)
+
+        self.to_out = nn.Sequential(
+            nn.Linear(inner_dim, dim),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x, attn_mask = None, memories = None):
+        x = self.norm(x)
+
+        x_kv = x # input for key / values projection
+
+        if exists(memories):
+            # add memories to key / values if it is passed in
+            memories = repeat(memories, 'n d -> b n d', b = x.shape[0]) if memories.ndim == 2 else memories
+            x_kv = torch.cat((x_kv, memories), dim = 1)
+
+        qkv = (self.to_q(x), *self.to_kv(x_kv).chunk(2, dim = -1))
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h = self.heads), qkv)
+
+        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
+        if exists(attn_mask):
+            dots = dots.masked_fill(~attn_mask, -torch.finfo(dots.dtype).max)
+
+        attn = self.attend(dots)
+        attn = self.dropout(attn)
+
+        out = torch.matmul(attn, v)
+        out = rearrange(out, 'b h n d -> b n (h d)')
+        return self.to_out(out)
+
+class Transformer(nn.Module):
+    def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout = 0.):
+        super().__init__()
+        self.layers = nn.ModuleList([])
+        for _ in range(depth):
+            self.layers.append(nn.ModuleList([
+                Attention(dim, heads = heads, dim_head = dim_head, dropout = dropout),
+                FeedForward(dim, mlp_dim, dropout = dropout)
+            ]))
+
+    def forward(self, x, attn_mask = None, memories = None):
+        for ind, (attn, ff) in enumerate(self.layers):
+            layer_memories = memories[ind] if exists(memories) else None
+
+            x = attn(x, attn_mask = attn_mask, memories = layer_memories) + x
+            x = ff(x) + x
+        return x
+
+class ViT(nn.Module):
+    def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, pool = 'cls', channels = 3, dim_head = 64, dropout = 0., emb_dropout = 0.):
+        super().__init__()
+        image_height, image_width = pair(image_size)
+        patch_height, patch_width = pair(patch_size)
+
+        assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'
+
+        num_patches = (image_height // patch_height) * (image_width // patch_width)
+        patch_dim = channels * patch_height * patch_width
+        assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)'
+
+        self.to_patch_embedding = nn.Sequential(
+            Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = patch_height, p2 = patch_width),
+            nn.Linear(patch_dim, dim),
+        )
+
+        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim))
+        self.cls_token = nn.Parameter(torch.randn(1, 1, dim))
+        self.dropout = nn.Dropout(emb_dropout)
+
+        self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, dropout)
+
+        self.mlp_head = nn.Sequential(
+            nn.LayerNorm(dim),
+            nn.Linear(dim, num_classes)
+        )
+
+    def img_to_tokens(self, img):
+        x = self.to_patch_embedding(img)
+
+        cls_tokens = repeat(self.cls_token, '1 n d -> b n d', b = x.shape[0])
+        x = torch.cat((cls_tokens, x), dim = 1)
+
+        x += self.pos_embedding
+        x = self.dropout(x)
+        return x
+
+    def forward(self, img):
+        x = self.img_to_tokens(img)        
+
+        x = self.transformer(x)
+
+        cls_tokens = x[:, 0]
+        return self.mlp_head(cls_tokens)
+
+# adapter with learnable memories per layer, memory CLS token, and learnable adapter head
+
+class Adapter(nn.Module):
+    def __init__(
+        self,
+        *,
+        vit,
+        num_memories_per_layer = 10,
+        num_classes = 2,   
+    ):
+        super().__init__()
+        assert isinstance(vit, ViT)
+
+        # extract some model variables needed
+
+        dim = vit.cls_token.shape[-1]
+        layers = len(vit.transformer.layers)
+        num_patches = vit.pos_embedding.shape[-2]
+
+        self.vit = vit
+
+        # freeze ViT backbone - only memories will be finetuned
+
+        freeze_all_layers_(vit)
+
+        # learnable parameters
+
+        self.memory_cls_token = nn.Parameter(torch.randn(dim))
+        self.memories_per_layer = nn.Parameter(torch.randn(layers, num_memories_per_layer, dim))
+
+        self.mlp_head = nn.Sequential(
+            nn.LayerNorm(dim),
+            nn.Linear(dim, num_classes)
+        )
+
+        # specialized attention mask to preserve the output of the original ViT
+        # it allows the memory CLS token to attend to all other tokens (and the learnable memory layer tokens), but not vice versa        
+
+        attn_mask = torch.ones((num_patches, num_patches), dtype = torch.bool)
+        attn_mask = F.pad(attn_mask, (1, num_memories_per_layer), value = False)  # main tokens cannot attend to learnable memories per layer
+        attn_mask = F.pad(attn_mask, (0, 0, 1, 0), value = True)                  # memory CLS token can attend to everything
+        self.register_buffer('attn_mask', attn_mask)
+
+    def forward(self, img):
+        b = img.shape[0]
+
+        tokens = self.vit.img_to_tokens(img)
+
+        # add task specific memory tokens
+
+        memory_cls_tokens = repeat(self.memory_cls_token, 'd -> b 1 d', b = b)
+        tokens = torch.cat((memory_cls_tokens, tokens), dim = 1)        
+
+        # pass memories along with image tokens through transformer for attending
+
+        out = self.vit.transformer(tokens, memories = self.memories_per_layer, attn_mask = self.attn_mask)
+
+        # extract memory CLS tokens
+
+        memory_cls_tokens = out[:, 0]
+
+        # pass through task specific adapter head
+
+        return self.mlp_head(memory_cls_tokens)

vit_pytorch/levit.py

@@ -52,6 +52,7 @@ class Attention(nn.Module):
         self.to_v = nn.Sequential(nn.Conv2d(dim, inner_dim_value, 1, bias = False), nn.BatchNorm2d(inner_dim_value))
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
 
         out_batch_norm = nn.BatchNorm2d(dim_out)
         nn.init.zeros_(out_batch_norm.weight)
@@ -100,6 +101,7 @@ class Attention(nn.Module):
         dots = self.apply_pos_bias(dots)
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)
         out = rearrange(out, 'b h (x y) d -> b (h d) x y', h = h, y = y)

vit_pytorch/local_vit.py

@@ -78,6 +78,7 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
         self.to_out = nn.Sequential(
@@ -93,6 +94,7 @@ class Attention(nn.Module):
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')

vit_pytorch/mobile_vit.py

@@ -54,6 +54,8 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim=-1)
+        self.dropout = nn.Dropout(dropout)
+
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias=False)
 
         self.to_out = nn.Sequential(
@@ -67,7 +69,10 @@ class Attention(nn.Module):
             t, 'b p n (h d) -> b p h n d', h=self.heads), qkv)
 
         dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
         attn = self.attend(dots)
+        attn = self.dropout(attn)
+
         out = torch.matmul(attn, v)
         out = rearrange(out, 'b p h n d -> b p n (h d)')
         return self.to_out(out)

vit_pytorch/nest.py

@@ -55,6 +55,7 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
         self.to_qkv = nn.Conv2d(dim, inner_dim * 3, 1, bias = False)
 
         self.to_out = nn.Sequential(
@@ -71,6 +72,7 @@ class Attention(nn.Module):
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)
         out = rearrange(out, 'b h (x y) d -> b (h d) x y', x = h, y = w)

vit_pytorch/parallel_vit.py

@@ -0,0 +1,140 @@
+import torch
+from torch import nn
+
+from einops import rearrange, repeat
+from einops.layers.torch import Rearrange
+
+# helpers
+
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
+# classes
+
+class Parallel(nn.Module):
+    def __init__(self, *fns):
+        super().__init__()
+        self.fns = nn.ModuleList(fns)
+
+    def forward(self, x):
+        return sum([fn(x) for fn in self.fns])
+
+class PreNorm(nn.Module):
+    def __init__(self, dim, fn):
+        super().__init__()
+        self.norm = nn.LayerNorm(dim)
+        self.fn = fn
+    def forward(self, x, **kwargs):
+        return self.fn(self.norm(x), **kwargs)
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, hidden_dim, dropout = 0.):
+        super().__init__()
+        self.net = nn.Sequential(
+            nn.Linear(dim, hidden_dim),
+            nn.GELU(),
+            nn.Dropout(dropout),
+            nn.Linear(hidden_dim, dim),
+            nn.Dropout(dropout)
+        )
+    def forward(self, x):
+        return self.net(x)
+
+class Attention(nn.Module):
+    def __init__(self, dim, heads = 8, dim_head = 64, dropout = 0.):
+        super().__init__()
+        inner_dim = dim_head *  heads
+        project_out = not (heads == 1 and dim_head == dim)
+
+        self.heads = heads
+        self.scale = dim_head ** -0.5
+
+        self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
+        self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
+
+        self.to_out = nn.Sequential(
+            nn.Linear(inner_dim, dim),
+            nn.Dropout(dropout)
+        ) if project_out else nn.Identity()
+
+    def forward(self, x):
+        qkv = self.to_qkv(x).chunk(3, dim = -1)
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h = self.heads), qkv)
+
+        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
+        attn = self.attend(dots)
+        attn = self.dropout(attn)
+
+        out = torch.matmul(attn, v)
+        out = rearrange(out, 'b h n d -> b n (h d)')
+        return self.to_out(out)
+
+class Transformer(nn.Module):
+    def __init__(self, dim, depth, heads, dim_head, mlp_dim, num_parallel_branches = 2, dropout = 0.):
+        super().__init__()
+        self.layers = nn.ModuleList([])
+
+        attn_block = lambda: PreNorm(dim, Attention(dim, heads = heads, dim_head = dim_head, dropout = dropout))
+        ff_block = lambda: PreNorm(dim, FeedForward(dim, mlp_dim, dropout = dropout))        
+
+        for _ in range(depth):
+            self.layers.append(nn.ModuleList([
+                Parallel(*[attn_block() for _ in range(num_parallel_branches)]),
+                Parallel(*[ff_block() for _ in range(num_parallel_branches)]),
+            ]))
+
+    def forward(self, x):
+        for attns, ffs in self.layers:
+            x = attns(x) + x
+            x = ffs(x) + x
+        return x
+
+class ViT(nn.Module):
+    def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, pool = 'cls', num_parallel_branches = 2, channels = 3, dim_head = 64, dropout = 0., emb_dropout = 0.):
+        super().__init__()
+        image_height, image_width = pair(image_size)
+        patch_height, patch_width = pair(patch_size)
+
+        assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'
+
+        num_patches = (image_height // patch_height) * (image_width // patch_width)
+        patch_dim = channels * patch_height * patch_width
+        assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)'
+
+        self.to_patch_embedding = nn.Sequential(
+            Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = patch_height, p2 = patch_width),
+            nn.Linear(patch_dim, dim),
+        )
+
+        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim))
+        self.cls_token = nn.Parameter(torch.randn(1, 1, dim))
+        self.dropout = nn.Dropout(emb_dropout)
+
+        self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, num_parallel_branches, dropout)
+
+        self.pool = pool
+        self.to_latent = nn.Identity()
+
+        self.mlp_head = nn.Sequential(
+            nn.LayerNorm(dim),
+            nn.Linear(dim, num_classes)
+        )
+
+    def forward(self, img):
+        x = self.to_patch_embedding(img)
+        b, n, _ = x.shape
+
+        cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
+        x = torch.cat((cls_tokens, x), dim=1)
+        x += self.pos_embedding[:, :(n + 1)]
+        x = self.dropout(x)
+
+        x = self.transformer(x)
+
+        x = x.mean(dim = 1) if self.pool == 'mean' else x[:, 0]
+
+        x = self.to_latent(x)
+        return self.mlp_head(x)

vit_pytorch/pit.py

@@ -48,6 +48,7 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
         self.to_out = nn.Sequential(
@@ -63,6 +64,7 @@ class Attention(nn.Module):
         dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')

vit_pytorch/regionvit.py

@@ -61,8 +61,13 @@ class Attention(nn.Module):
         inner_dim = dim_head * heads
 
         self.norm = nn.LayerNorm(dim)
+        self.dropout = nn.Dropout(dropout)
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
-        self.to_out = nn.Linear(inner_dim, dim)
+
+        self.to_out = nn.Sequential(
+            nn.Linear(inner_dim, dim),
+            nn.Dropout(dropout)
+        )
 
     def forward(self, x, rel_pos_bias = None):
         h = self.heads
@@ -86,6 +91,7 @@ class Attention(nn.Module):
             sim = sim + rel_pos_bias
 
         attn = sim.softmax(dim = -1)
+        attn = self.dropout(attn)
 
         # merge heads
 

vit_pytorch/rvt.py

@@ -104,6 +104,7 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
 
         self.use_ds_conv = use_ds_conv
 
@@ -148,6 +149,7 @@ class Attention(nn.Module):
         dots = einsum('b i d, b j d -> b i j', q, k) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = einsum('b i j, b j d -> b i d', attn, v)
         out = rearrange(out, '(b h) n d -> b n (h d)', h = h)

vit_pytorch/scalable_vit.py

@@ -0,0 +1,306 @@
+from functools import partial
+import torch
+from torch import nn
+
+from einops import rearrange, repeat
+from einops.layers.torch import Rearrange, Reduce
+
+# helpers
+
+def exists(val):
+    return val is not None
+
+def default(val, d):
+    return val if exists(val) else d
+
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
+def cast_tuple(val, length = 1):
+    return val if isinstance(val, tuple) else ((val,) * length)
+
+# helper classes
+
+class ChanLayerNorm(nn.Module):
+    def __init__(self, dim, eps = 1e-5):
+        super().__init__()
+        self.eps = eps
+        self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
+        self.b = nn.Parameter(torch.zeros(1, dim, 1, 1))
+
+    def forward(self, x):
+        var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
+        mean = torch.mean(x, dim = 1, keepdim = True)
+        return (x - mean) / (var + self.eps).sqrt() * self.g + self.b
+
+class PreNorm(nn.Module):
+    def __init__(self, dim, fn):
+        super().__init__()
+        self.norm = ChanLayerNorm(dim)
+        self.fn = fn
+
+    def forward(self, x):
+        return self.fn(self.norm(x))
+
+class Downsample(nn.Module):
+    def __init__(self, dim_in, dim_out):
+        super().__init__()
+        self.conv = nn.Conv2d(dim_in, dim_out, 3, stride = 2, padding = 1)
+
+    def forward(self, x):
+        return self.conv(x)
+
+class PEG(nn.Module):
+    def __init__(self, dim, kernel_size = 3):
+        super().__init__()
+        self.proj = nn.Conv2d(dim, dim, kernel_size = kernel_size, padding = kernel_size // 2, groups = dim, stride = 1)
+
+    def forward(self, x):
+        return self.proj(x) + x
+
+# feedforward
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, expansion_factor = 4, dropout = 0.):
+        super().__init__()
+        inner_dim = dim * expansion_factor
+        self.net = nn.Sequential(
+            nn.Conv2d(dim, inner_dim, 1),
+            nn.GELU(),
+            nn.Dropout(dropout),
+            nn.Conv2d(inner_dim, dim, 1),
+            nn.Dropout(dropout)
+        )
+    def forward(self, x):
+        return self.net(x)
+
+# attention
+
+class ScalableSelfAttention(nn.Module):
+    def __init__(
+        self,
+        dim,
+        heads = 8,
+        dim_key = 32,
+        dim_value = 32,
+        dropout = 0.,
+        reduction_factor = 1
+    ):
+        super().__init__()
+        self.heads = heads
+        self.scale = dim_key ** -0.5
+        self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
+        self.to_q = nn.Conv2d(dim, dim_key * heads, 1, bias = False)
+        self.to_k = nn.Conv2d(dim, dim_key * heads, reduction_factor, stride = reduction_factor, bias = False)
+        self.to_v = nn.Conv2d(dim, dim_value * heads, reduction_factor, stride = reduction_factor, bias = False)
+
+        self.to_out = nn.Sequential(
+            nn.Conv2d(dim_value * heads, dim, 1),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x):
+        height, width, heads = *x.shape[-2:], self.heads
+
+        q, k, v = self.to_q(x), self.to_k(x), self.to_v(x)
+
+        # split out heads
+
+        q, k, v = map(lambda t: rearrange(t, 'b (h d) ... -> b h (...) d', h = heads), (q, k, v))
+
+        # similarity
+
+        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
+        # attention
+
+        attn = self.attend(dots)
+        attn = self.dropout(attn)
+
+        # aggregate values
+
+        out = torch.matmul(attn, v)
+
+        # merge back heads
+
+        out = rearrange(out, 'b h (x y) d -> b (h d) x y', x = height, y = width)
+        return self.to_out(out)
+
+class InteractiveWindowedSelfAttention(nn.Module):
+    def __init__(
+        self,
+        dim,
+        window_size,
+        heads = 8,
+        dim_key = 32,
+        dim_value = 32,
+        dropout = 0.
+    ):
+        super().__init__()
+        self.heads = heads
+        self.scale = dim_key ** -0.5
+        self.window_size = window_size
+        self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
+        self.local_interactive_module = nn.Conv2d(dim_value * heads, dim_value * heads, 3, padding = 1)
+
+        self.to_q = nn.Conv2d(dim, dim_key * heads, 1, bias = False)
+        self.to_k = nn.Conv2d(dim, dim_key * heads, 1, bias = False)
+        self.to_v = nn.Conv2d(dim, dim_value * heads, 1, bias = False)
+
+        self.to_out = nn.Sequential(
+            nn.Conv2d(dim_value * heads, dim, 1),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x):
+        height, width, heads, wsz = *x.shape[-2:], self.heads, self.window_size
+
+        wsz_h, wsz_w = default(wsz, height), default(wsz, width)
+        assert (height % wsz_h) == 0 and (width % wsz_w) == 0, f'height ({height}) or width ({width}) of feature map is not divisible by the window size ({wsz_h}, {wsz_w})'
+
+        q, k, v = self.to_q(x), self.to_k(x), self.to_v(x)
+
+        # get output of LIM
+
+        local_out = self.local_interactive_module(v)
+
+        # divide into window (and split out heads) for efficient self attention
+
+        q, k, v = map(lambda t: rearrange(t, 'b (h d) (x w1) (y w2) -> (b x y) h (w1 w2) d', h = heads, w1 = wsz_h, w2 = wsz_w), (q, k, v))
+
+        # similarity
+
+        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
+        # attention
+
+        attn = self.attend(dots)
+        attn = self.dropout(attn)
+
+        # aggregate values
+
+        out = torch.matmul(attn, v)
+
+        # reshape the windows back to full feature map (and merge heads)
+
+        out = rearrange(out, '(b x y) h (w1 w2) d -> b (h d) (x w1) (y w2)', x = height // wsz_h, y = width // wsz_w, w1 = wsz_h, w2 = wsz_w)
+
+        # add LIM output 
+
+        out = out + local_out
+
+        return self.to_out(out)
+
+class Transformer(nn.Module):
+    def __init__(
+        self,
+        dim,
+        depth,
+        heads = 8,
+        ff_expansion_factor = 4,
+        dropout = 0.,
+        ssa_dim_key = 32,
+        ssa_dim_value = 32,
+        ssa_reduction_factor = 1,
+        iwsa_dim_key = 32,
+        iwsa_dim_value = 32,
+        iwsa_window_size = None,
+        norm_output = True
+    ):
+        super().__init__()
+        self.layers = nn.ModuleList([])
+        for ind in range(depth):
+            is_first = ind == 0
+
+            self.layers.append(nn.ModuleList([
+                PreNorm(dim, ScalableSelfAttention(dim, heads = heads, dim_key = ssa_dim_key, dim_value = ssa_dim_value, reduction_factor = ssa_reduction_factor, dropout = dropout)),
+                PreNorm(dim, FeedForward(dim, expansion_factor = ff_expansion_factor, dropout = dropout)),
+                PEG(dim) if is_first else None,
+                PreNorm(dim, FeedForward(dim, expansion_factor = ff_expansion_factor, dropout = dropout)),
+                PreNorm(dim, InteractiveWindowedSelfAttention(dim, heads = heads, dim_key = iwsa_dim_key, dim_value = iwsa_dim_value, window_size = iwsa_window_size, dropout = dropout))
+            ]))
+
+        self.norm = ChanLayerNorm(dim) if norm_output else nn.Identity()
+
+    def forward(self, x):
+        for ssa, ff1, peg, iwsa, ff2 in self.layers:
+            x = ssa(x) + x
+            x = ff1(x) + x
+
+            if exists(peg):
+                x = peg(x)
+
+            x = iwsa(x) + x
+            x = ff2(x) + x
+
+        return self.norm(x)
+
+class ScalableViT(nn.Module):
+    def __init__(
+        self,
+        *,
+        num_classes,
+        dim,
+        depth,
+        heads,
+        reduction_factor,
+        window_size = None,
+        iwsa_dim_key = 32,
+        iwsa_dim_value = 32,
+        ssa_dim_key = 32,
+        ssa_dim_value = 32,
+        ff_expansion_factor = 4,
+        channels = 3,
+        dropout = 0.
+    ):
+        super().__init__()
+        self.to_patches = nn.Conv2d(channels, dim, 7, stride = 4, padding = 3)
+
+        assert isinstance(depth, tuple), 'depth needs to be tuple if integers indicating number of transformer blocks at that stage'
+
+        num_stages = len(depth)
+        dims = tuple(map(lambda i: (2 ** i) * dim, range(num_stages)))
+
+        hyperparams_per_stage = [
+            heads,
+            ssa_dim_key,
+            ssa_dim_value,
+            reduction_factor,
+            iwsa_dim_key,
+            iwsa_dim_value,
+            window_size,
+        ]
+
+        hyperparams_per_stage = list(map(partial(cast_tuple, length = num_stages), hyperparams_per_stage))
+        assert all(tuple(map(lambda arr: len(arr) == num_stages, hyperparams_per_stage)))
+
+        self.layers = nn.ModuleList([])
+
+        for ind, (layer_dim, layer_depth, layer_heads, layer_ssa_dim_key, layer_ssa_dim_value, layer_ssa_reduction_factor, layer_iwsa_dim_key, layer_iwsa_dim_value, layer_window_size) in enumerate(zip(dims, depth, *hyperparams_per_stage)):
+            is_last = ind == (num_stages - 1)
+
+            self.layers.append(nn.ModuleList([
+                Transformer(dim = layer_dim, depth = layer_depth, heads = layer_heads, ff_expansion_factor = ff_expansion_factor, dropout = dropout, ssa_dim_key = layer_ssa_dim_key, ssa_dim_value = layer_ssa_dim_value, ssa_reduction_factor = layer_ssa_reduction_factor, iwsa_dim_key = layer_iwsa_dim_key, iwsa_dim_value = layer_iwsa_dim_value, iwsa_window_size = layer_window_size, norm_output = not is_last),
+                Downsample(layer_dim, layer_dim * 2) if not is_last else None
+            ]))
+
+        self.mlp_head = nn.Sequential(
+            Reduce('b d h w -> b d', 'mean'),
+            nn.LayerNorm(dims[-1]),
+            nn.Linear(dims[-1], num_classes)
+        )
+
+    def forward(self, img):
+        x = self.to_patches(img)
+
+        for transformer, downsample in self.layers:
+            x = transformer(x)
+
+            if exists(downsample):
+                x = downsample(x)
+
+        return self.mlp_head(x)

vit_pytorch/sep_vit.py

@@ -0,0 +1,296 @@
+from functools import partial
+
+import torch
+from torch import nn, einsum
+
+from einops import rearrange, repeat
+from einops.layers.torch import Rearrange, Reduce
+
+# helpers
+
+def cast_tuple(val, length = 1):
+    return val if isinstance(val, tuple) else ((val,) * length)
+
+# helper classes
+
+class ChanLayerNorm(nn.Module):
+    def __init__(self, dim, eps = 1e-5):
+        super().__init__()
+        self.eps = eps
+        self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
+        self.b = nn.Parameter(torch.zeros(1, dim, 1, 1))
+
+    def forward(self, x):
+        var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
+        mean = torch.mean(x, dim = 1, keepdim = True)
+        return (x - mean) / (var + self.eps).sqrt() * self.g + self.b
+
+class PreNorm(nn.Module):
+    def __init__(self, dim, fn):
+        super().__init__()
+        self.norm = ChanLayerNorm(dim)
+        self.fn = fn
+
+    def forward(self, x):
+        return self.fn(self.norm(x))
+
+class OverlappingPatchEmbed(nn.Module):
+    def __init__(self, dim_in, dim_out, stride = 2):
+        super().__init__()
+        kernel_size = stride * 2 - 1
+        padding = kernel_size // 2
+        self.conv = nn.Conv2d(dim_in, dim_out, kernel_size, stride = stride, padding = padding)
+
+    def forward(self, x):
+        return self.conv(x)
+
+class PEG(nn.Module):
+    def __init__(self, dim, kernel_size = 3):
+        super().__init__()
+        self.proj = nn.Conv2d(dim, dim, kernel_size = kernel_size, padding = kernel_size // 2, groups = dim, stride = 1)
+
+    def forward(self, x):
+        return self.proj(x) + x
+
+# feedforward
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, mult = 4, dropout = 0.):
+        super().__init__()
+        inner_dim = int(dim * mult)
+        self.net = nn.Sequential(
+            nn.Conv2d(dim, inner_dim, 1),
+            nn.GELU(),
+            nn.Dropout(dropout),
+            nn.Conv2d(inner_dim, dim, 1),
+            nn.Dropout(dropout)
+        )
+    def forward(self, x):
+        return self.net(x)
+
+# attention
+
+class DSSA(nn.Module):
+    def __init__(
+        self,
+        dim,
+        heads = 8,
+        dim_head = 32,
+        dropout = 0.,
+        window_size = 7
+    ):
+        super().__init__()
+        self.heads = heads
+        self.scale = dim_head ** -0.5
+        self.window_size = window_size
+        inner_dim = dim_head * heads
+
+        self.attend = nn.Sequential(
+            nn.Softmax(dim = -1),
+            nn.Dropout(dropout)
+        )
+
+        self.to_qkv = nn.Conv1d(dim, inner_dim * 3, 1, bias = False)
+
+        # window tokens
+
+        self.window_tokens = nn.Parameter(torch.randn(dim))
+
+        # prenorm and non-linearity for window tokens
+        # then projection to queries and keys for window tokens
+
+        self.window_tokens_to_qk = nn.Sequential(
+            nn.LayerNorm(dim_head),
+            nn.GELU(),
+            Rearrange('b h n c -> b (h c) n'),
+            nn.Conv1d(inner_dim, inner_dim * 2, 1, groups = heads),
+            Rearrange('b (h c) n -> b h n c', h = heads),
+        )
+
+        # window attention
+
+        self.window_attend = nn.Sequential(
+            nn.Softmax(dim = -1),
+            nn.Dropout(dropout)
+        )
+
+        self.to_out = nn.Sequential(
+            nn.Conv2d(inner_dim, dim, 1),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x):
+        """
+        einstein notation
+
+        b - batch
+        c - channels
+        w1 - window size (height)
+        w2 - also window size (width)
+        i - sequence dimension (source)
+        j - sequence dimension (target dimension to be reduced)
+        h - heads
+        x - height of feature map divided by window size
+        y - width of feature map divided by window size
+        """
+
+        batch, height, width, heads, wsz = x.shape[0], *x.shape[-2:], self.heads, self.window_size
+        assert (height % wsz) == 0 and (width % wsz) == 0, f'height {height} and width {width} must be divisible by window size {wsz}'
+        num_windows = (height // wsz) * (width // wsz)
+
+        # fold in windows for "depthwise" attention - not sure why it is named depthwise when it is just "windowed" attention
+
+        x = rearrange(x, 'b c (h w1) (w w2) -> (b h w) c (w1 w2)', w1 = wsz, w2 = wsz)
+
+        # add windowing tokens
+
+        w = repeat(self.window_tokens, 'c -> b c 1', b = x.shape[0])
+        x = torch.cat((w, x), dim = -1)
+
+        # project for queries, keys, value
+
+        q, k, v = self.to_qkv(x).chunk(3, dim = 1)
+
+        # split out heads
+
+        q, k, v = map(lambda t: rearrange(t, 'b (h d) ... -> b h (...) d', h = heads), (q, k, v))
+
+        # scale
+
+        q = q * self.scale
+
+        # similarity
+
+        dots = einsum('b h i d, b h j d -> b h i j', q, k)
+
+        # attention
+
+        attn = self.attend(dots)
+
+        # aggregate values
+
+        out = torch.matmul(attn, v)
+
+        # split out windowed tokens
+
+        window_tokens, windowed_fmaps = out[:, :, 0], out[:, :, 1:]
+
+        # early return if there is only 1 window
+
+        if num_windows == 1:
+            fmap = rearrange(windowed_fmaps, '(b x y) h (w1 w2) d -> b (h d) (x w1) (y w2)', x = height // wsz, y = width // wsz, w1 = wsz, w2 = wsz)
+            return self.to_out(fmap)
+
+        # carry out the pointwise attention, the main novelty in the paper
+
+        window_tokens = rearrange(window_tokens, '(b x y) h d -> b h (x y) d', x = height // wsz, y = width // wsz)
+        windowed_fmaps = rearrange(windowed_fmaps, '(b x y) h n d -> b h (x y) n d', x = height // wsz, y = width // wsz)
+
+        # windowed queries and keys (preceded by prenorm activation)
+
+        w_q, w_k = self.window_tokens_to_qk(window_tokens).chunk(2, dim = -1)
+
+        # scale
+
+        w_q = w_q * self.scale
+
+        # similarities
+
+        w_dots = einsum('b h i d, b h j d -> b h i j', w_q, w_k)
+
+        w_attn = self.window_attend(w_dots)
+
+        # aggregate the feature maps from the "depthwise" attention step (the most interesting part of the paper, one i haven't seen before)
+
+        aggregated_windowed_fmap = einsum('b h i j, b h j w d -> b h i w d', w_attn, windowed_fmaps)
+
+        # fold back the windows and then combine heads for aggregation
+
+        fmap = rearrange(aggregated_windowed_fmap, 'b h (x y) (w1 w2) d -> b (h d) (x w1) (y w2)', x = height // wsz, y = width // wsz, w1 = wsz, w2 = wsz)
+        return self.to_out(fmap)
+
+class Transformer(nn.Module):
+    def __init__(
+        self,
+        dim,
+        depth,
+        dim_head = 32,
+        heads = 8,
+        ff_mult = 4,
+        dropout = 0.,
+        norm_output = True
+    ):
+        super().__init__()
+        self.layers = nn.ModuleList([])
+
+        for ind in range(depth):
+            self.layers.append(nn.ModuleList([
+                PreNorm(dim, DSSA(dim, heads = heads, dim_head = dim_head, dropout = dropout)),
+                PreNorm(dim, FeedForward(dim, mult = ff_mult, dropout = dropout)),
+            ]))
+
+        self.norm = ChanLayerNorm(dim) if norm_output else nn.Identity()
+
+    def forward(self, x):
+        for attn, ff in self.layers:
+            x = attn(x) + x
+            x = ff(x) + x
+
+        return self.norm(x)
+
+class SepViT(nn.Module):
+    def __init__(
+        self,
+        *,
+        num_classes,
+        dim,
+        depth,
+        heads,
+        window_size = 7,
+        dim_head = 32,
+        ff_mult = 4,
+        channels = 3,
+        dropout = 0.
+    ):
+        super().__init__()
+        self.to_patches = nn.Conv2d(channels, dim, 7, stride = 4, padding = 3)
+
+        assert isinstance(depth, tuple), 'depth needs to be tuple if integers indicating number of transformer blocks at that stage'
+
+        num_stages = len(depth)
+
+        dims = tuple(map(lambda i: (2 ** i) * dim, range(num_stages)))
+        dims = (channels, *dims)
+        dim_pairs = tuple(zip(dims[:-1], dims[1:]))
+
+        strides = (4, *((2,) * (num_stages - 1)))
+
+        hyperparams_per_stage = [heads, window_size]
+        hyperparams_per_stage = list(map(partial(cast_tuple, length = num_stages), hyperparams_per_stage))
+        assert all(tuple(map(lambda arr: len(arr) == num_stages, hyperparams_per_stage)))
+
+        self.layers = nn.ModuleList([])
+
+        for ind, ((layer_dim_in, layer_dim), layer_depth, layer_stride, layer_heads, layer_window_size) in enumerate(zip(dim_pairs, depth, strides, *hyperparams_per_stage)):
+            is_last = ind == (num_stages - 1)
+
+            self.layers.append(nn.ModuleList([
+                OverlappingPatchEmbed(layer_dim_in, layer_dim, stride = layer_stride),
+                PEG(layer_dim),
+                Transformer(dim = layer_dim, depth = layer_depth, heads = layer_heads, ff_mult = ff_mult, dropout = dropout, norm_output = not is_last),
+            ]))
+
+        self.mlp_head = nn.Sequential(
+            Reduce('b d h w -> b d', 'mean'),
+            nn.LayerNorm(dims[-1]),
+            nn.Linear(dims[-1], num_classes)
+        )
+
+    def forward(self, x):
+
+        for ope, peg, transformer in self.layers:
+            x = ope(x)
+            x = peg(x)
+            x = transformer(x)
+
+        return self.mlp_head(x)

vit_pytorch/twins_svt.py

@@ -130,6 +130,8 @@ class GlobalAttention(nn.Module):
         self.to_q = nn.Conv2d(dim, inner_dim, 1, bias = False)
         self.to_kv = nn.Conv2d(dim, inner_dim * 2, k, stride = k, bias = False)
 
+        self.dropout = nn.Dropout(dropout)
+
         self.to_out = nn.Sequential(
             nn.Conv2d(inner_dim, dim, 1),
             nn.Dropout(dropout)
@@ -145,6 +147,7 @@ class GlobalAttention(nn.Module):
         dots = einsum('b i d, b j d -> b i j', q, k) * self.scale
 
         attn = dots.softmax(dim = -1)
+        attn = self.dropout(attn)
 
         out = einsum('b i j, b j d -> b i d', attn, v)
         out = rearrange(out, '(b h) (x y) d -> b (h d) x y', h = h, y = y)

vit_pytorch/vit.py

@@ -42,6 +42,8 @@ class Attention(nn.Module):
         self.scale = dim_head ** -0.5
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
         self.to_out = nn.Sequential(
@@ -56,6 +58,7 @@ class Attention(nn.Module):
         dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = torch.matmul(attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')
@@ -111,7 +114,7 @@ class ViT(nn.Module):
         x = self.to_patch_embedding(img)
         b, n, _ = x.shape
 
-        cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
+        cls_tokens = repeat(self.cls_token, '1 n d -> b n d', b = b)
         x = torch.cat((cls_tokens, x), dim=1)
         x += self.pos_embedding[:, :(n + 1)]
         x = self.dropout(x)

vit_pytorch/vit_for_small_dataset.py

@@ -42,6 +42,8 @@ class LSA(nn.Module):
         self.temperature = nn.Parameter(torch.log(torch.tensor(dim_head ** -0.5)))
 
         self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
         self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
 
         self.to_out = nn.Sequential(
@@ -60,6 +62,7 @@ class LSA(nn.Module):
         dots = dots.masked_fill(mask, mask_value)
 
         attn = self.attend(dots)
+        attn = self.dropout(attn)
 
         out = torch.matmul(attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')

vit_pytorch/vit_with_patch_merger.py

@@ -0,0 +1,147 @@
+import torch
+from torch import nn
+
+from einops import rearrange, repeat
+from einops.layers.torch import Rearrange, Reduce
+
+# helpers
+
+def exists(val):
+    return val is not None
+
+def default(val ,d):
+    return val if exists(val) else d
+
+def pair(t):
+    return t if isinstance(t, tuple) else (t, t)
+
+# patch merger class
+
+class PatchMerger(nn.Module):
+    def __init__(self, dim, num_tokens_out):
+        super().__init__()
+        self.scale = dim ** -0.5
+        self.norm = nn.LayerNorm(dim)
+        self.queries = nn.Parameter(torch.randn(num_tokens_out, dim))
+
+    def forward(self, x):
+        x = self.norm(x)
+        sim = torch.matmul(self.queries, x.transpose(-1, -2)) * self.scale
+        attn = sim.softmax(dim = -1)
+        return torch.matmul(attn, x)
+
+# classes
+
+class PreNorm(nn.Module):
+    def __init__(self, dim, fn):
+        super().__init__()
+        self.norm = nn.LayerNorm(dim)
+        self.fn = fn
+    def forward(self, x, **kwargs):
+        return self.fn(self.norm(x), **kwargs)
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, hidden_dim, dropout = 0.):
+        super().__init__()
+        self.net = nn.Sequential(
+            nn.Linear(dim, hidden_dim),
+            nn.GELU(),
+            nn.Dropout(dropout),
+            nn.Linear(hidden_dim, dim),
+            nn.Dropout(dropout)
+        )
+    def forward(self, x):
+        return self.net(x)
+
+class Attention(nn.Module):
+    def __init__(self, dim, heads = 8, dim_head = 64, dropout = 0.):
+        super().__init__()
+        inner_dim = dim_head *  heads
+        project_out = not (heads == 1 and dim_head == dim)
+
+        self.heads = heads
+        self.scale = dim_head ** -0.5
+
+        self.attend = nn.Softmax(dim = -1)
+        self.dropout = nn.Dropout(dropout)
+
+        self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
+
+        self.to_out = nn.Sequential(
+            nn.Linear(inner_dim, dim),
+            nn.Dropout(dropout)
+        ) if project_out else nn.Identity()
+
+    def forward(self, x):
+        qkv = self.to_qkv(x).chunk(3, dim = -1)
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h = self.heads), qkv)
+
+        dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
+
+        attn = self.attend(dots)
+        attn = self.dropout(attn)
+
+        out = torch.matmul(attn, v)
+        out = rearrange(out, 'b h n d -> b n (h d)')
+        return self.to_out(out)
+
+class Transformer(nn.Module):
+    def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout = 0., patch_merge_layer = None, patch_merge_num_tokens = 8):
+        super().__init__()
+        self.layers = nn.ModuleList([])
+
+        self.patch_merge_layer_index = default(patch_merge_layer, depth // 2) - 1 # default to mid-way through transformer, as shown in paper
+        self.patch_merger = PatchMerger(dim = dim, num_tokens_out = patch_merge_num_tokens)
+
+        for _ in range(depth):
+            self.layers.append(nn.ModuleList([
+                PreNorm(dim, Attention(dim, heads = heads, dim_head = dim_head, dropout = dropout)),
+                PreNorm(dim, FeedForward(dim, mlp_dim, dropout = dropout))
+            ]))
+    def forward(self, x):
+        for index, (attn, ff) in enumerate(self.layers):
+            x = attn(x) + x
+            x = ff(x) + x
+
+            if index == self.patch_merge_layer_index:
+                x = self.patch_merger(x)
+
+        return x
+
+class ViT(nn.Module):
+    def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, patch_merge_layer = None, patch_merge_num_tokens = 8, channels = 3, dim_head = 64, dropout = 0., emb_dropout = 0.):
+        super().__init__()
+        image_height, image_width = pair(image_size)
+        patch_height, patch_width = pair(patch_size)
+
+        assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'
+
+        num_patches = (image_height // patch_height) * (image_width // patch_width)
+        patch_dim = channels * patch_height * patch_width
+
+        self.to_patch_embedding = nn.Sequential(
+            Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = patch_height, p2 = patch_width),
+            nn.Linear(patch_dim, dim),
+        )
+
+        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim))
+        self.dropout = nn.Dropout(emb_dropout)
+
+        self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, dropout, patch_merge_layer, patch_merge_num_tokens)
+
+        self.mlp_head = nn.Sequential(
+            Reduce('b n d -> b d', 'mean'),
+            nn.LayerNorm(dim),
+            nn.Linear(dim, num_classes)
+        )
+
+    def forward(self, img):
+        x = self.to_patch_embedding(img)
+        b, n, _ = x.shape
+
+        x += self.pos_embedding[:, :n]
+        x = self.dropout(x)
+
+        x = self.transformer(x)
+
+        return self.mlp_head(x)