Skip to content

Commit

Permalink
add the 3d simple vit
Browse files Browse the repository at this point in the history
  • Loading branch information
lucidrains committed Oct 17, 2022
1 parent 29fbf0a commit b4853d3
Show file tree
Hide file tree
Showing 3 changed files with 153 additions and 2 deletions.
25 changes: 24 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -974,7 +974,7 @@ By popular request, I will start extending a few of the architectures in this re

You will need to pass in two additional hyperparameters: (1) the number of frames `frames` and (2) patch size along the frame dimension `frame_patch_size`

For starters, with the most basic ViT
For starters, 3D ViT

```python
import torch
Expand All @@ -999,6 +999,29 @@ video = torch.randn(4, 3, 16, 128, 128) # (batch, channels, frames, height, widt
preds = v(video) # (4, 1000)
```

3D Simple ViT

```python
import torch
from vit_pytorch.simple_vit_3d import SimpleViT

v = SimpleViT(
image_size = 128, # image size
frames = 16, # number of frames
image_patch_size = 16, # image patch size
frame_patch_size = 2, # frame patch size
num_classes = 1000,
dim = 1024,
depth = 6,
heads = 8,
mlp_dim = 2048
)

video = torch.randn(4, 3, 16, 128, 128) # (batch, channels, frames, height, width)

preds = v(video) # (4, 1000)
```

## Parallel ViT

<img src="./images/parallel-vit.png" width="350px"></img>
Expand Down
2 changes: 1 addition & 1 deletion setup.py
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
setup(
name = 'vit-pytorch',
packages = find_packages(exclude=['examples']),
version = '0.36.0',
version = '0.36.1',
license='MIT',
description = 'Vision Transformer (ViT) - Pytorch',
long_description_content_type = 'text/markdown',
Expand Down
128 changes: 128 additions & 0 deletions vit_pytorch/simple_vit_3d.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,128 @@
import torch
import torch.nn.functional as F
from torch import nn

from einops import rearrange
from einops.layers.torch import Rearrange

# helpers

def pair(t):
return t if isinstance(t, tuple) else (t, t)

def posemb_sincos_3d(patches, temperature = 10000, dtype = torch.float32):
_, f, h, w, dim, device, dtype = *patches.shape, patches.device, patches.dtype

z, y, x = torch.meshgrid(
torch.arange(f, device = device),
torch.arange(h, device = device),
torch.arange(w, device = device),
indexing = 'ij')

fourier_dim = dim // 6

omega = torch.arange(fourier_dim, device = device) / (fourier_dim - 1)
omega = 1. / (temperature ** omega)

z = z.flatten()[:, None] * omega[None, :]
y = y.flatten()[:, None] * omega[None, :]
x = x.flatten()[:, None] * omega[None, :]

pe = torch.cat((x.sin(), x.cos(), y.sin(), y.cos(), z.sin(), z.cos()), dim = 1)

pe = F.pad(pe, (0, dim - (fourier_dim * 6))) # pad if feature dimension not cleanly divisible by 6
return pe.type(dtype)

# classes

class FeedForward(nn.Module):
def __init__(self, dim, hidden_dim):
super().__init__()
self.net = nn.Sequential(
nn.LayerNorm(dim),
nn.Linear(dim, hidden_dim),
nn.GELU(),
nn.Linear(hidden_dim, dim),
)
def forward(self, x):
return self.net(x)

class Attention(nn.Module):
def __init__(self, dim, heads = 8, dim_head = 64):
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.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
self.to_out = nn.Linear(inner_dim, dim, bias = False)

def forward(self, x):
x = self.norm(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)

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):
super().__init__()
self.layers = nn.ModuleList([])
for _ in range(depth):
self.layers.append(nn.ModuleList([
Attention(dim, heads = heads, dim_head = dim_head),
FeedForward(dim, mlp_dim)
]))
def forward(self, x):
for attn, ff in self.layers:
x = attn(x) + x
x = ff(x) + x
return x

class SimpleViT(nn.Module):
def __init__(self, *, image_size, image_patch_size, frames, frame_patch_size, num_classes, dim, depth, heads, mlp_dim, channels = 3, dim_head = 64):
super().__init__()
image_height, image_width = pair(image_size)
patch_height, patch_width = pair(image_patch_size)

assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'
assert frames % frame_patch_size == 0, 'Frames must be divisible by the frame patch size'

num_patches = (image_height // patch_height) * (image_width // patch_width) * (frames // frame_patch_size)
patch_dim = channels * patch_height * patch_width * frame_patch_size

self.to_patch_embedding = nn.Sequential(
Rearrange('b c (f pf) (h p1) (w p2) -> b f h w (p1 p2 pf c)', p1 = patch_height, p2 = patch_width, pf = frame_patch_size),
nn.Linear(patch_dim, dim),
)

self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim)

self.to_latent = nn.Identity()
self.linear_head = nn.Sequential(
nn.LayerNorm(dim),
nn.Linear(dim, num_classes)
)

def forward(self, img):
*_, h, w, dtype = *img.shape, img.dtype

x = self.to_patch_embedding(img)
pe = posemb_sincos_3d(x)
x = rearrange(x, 'b ... d -> b (...) d') + pe

x = self.transformer(x)
x = x.mean(dim = 1)

x = self.to_latent(x)
return self.linear_head(x)

0 comments on commit b4853d3

Please sign in to comment.