| 288 | """ |
| 289 | |
| 290 | def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768, proj_padding=False): |
| 291 | super().__init__() |
| 292 | img_size = to_2tuple(img_size) |
| 293 | patch_size = to_2tuple(patch_size) |
| 294 | self.patch_shape = (img_size[0] // patch_size[0], img_size[1] // patch_size[1]) # could be dynamic |
| 295 | self.num_patches = self.patch_shape[0] * self.patch_shape[1] # could be dynamic |
| 296 | self.img_size = img_size |
| 297 | self.patch_size = patch_size |
| 298 | |
| 299 | if proj_padding: |
| 300 | self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size, padding=2) |
| 301 | else: |
| 302 | self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) |
| 303 | |
| 304 | def forward(self, x, **kwargs): |
| 305 | B, C, H, W = x.shape |