support rectangular images

Author: lucidrains

README.md

@@ -33,6 +33,25 @@ img = torch.randn(1, 3, 256, 256)
 pred = model(img) # (1, 1000)
 ```
 
+Rectangular image
+
+```python
+import torch
+from mlp_mixer_pytorch import MLPMixer
+
+model = MLPMixer(
+    image_size = (256, 128),
+    channels = 3,
+    patch_size = 16,
+    dim = 512,
+    depth = 12,
+    num_classes = 1000
+)
+
+img = torch.randn(1, 3, 256, 128)
+pred = model(img) # (1, 1000)
+```
+
 ## Citations
 
 ```bibtex

mlp_mixer_pytorch/mlp_mixer_pytorch.py

@@ -2,6 +2,8 @@
 from functools import partial
 from einops.layers.torch import Rearrange, Reduce
 
+pair = lambda x: x if isinstance(x, tuple) else (x, x)
+
 class PreNormResidual(nn.Module):
     def __init__(self, dim, fn):
         super().__init__()
@@ -22,8 +24,9 @@ def FeedForward(dim, expansion_factor = 4, dropout = 0., dense = nn.Linear):
     )
 
 def MLPMixer(*, image_size, channels, patch_size, dim, depth, num_classes, expansion_factor = 4, expansion_factor_token = 0.5, dropout = 0.):
-    assert (image_size % patch_size) == 0, 'image must be divisible by patch size'
-    num_patches = (image_size // patch_size) ** 2
+    image_h, image_w = pair(image_size)
+    assert (image_h % patch_size) == 0 and (image_w % patch_size) == 0, 'image must be divisible by patch size'
+    num_patches = (image_h // patch_size) * (image_w // patch_size)
     chan_first, chan_last = partial(nn.Conv1d, kernel_size = 1), nn.Linear
 
     return nn.Sequential(

setup.py

@@ -3,7 +3,7 @@
 setup(
   name = 'mlp-mixer-pytorch',
   packages = find_packages(exclude=[]),
-  version = '0.1.0',
+  version = '0.1.1',
   license='MIT',
   description = 'MLP Mixer - Pytorch',
   author = 'Phil Wang',