| 28 | """ |
| 29 | |
| 30 | def __init__( |
| 31 | self, |
| 32 | img_size: Union[int, Tuple[int, int]] = 224, |
| 33 | patch_size: Union[int, Tuple[int, int]] = 16, |
| 34 | in_chans: int = 3, |
| 35 | embed_dim: int = 768, |
| 36 | norm_layer: Optional[Callable] = None, |
| 37 | flatten_embedding: bool = True, |
| 38 | ) -> None: |
| 39 | super().__init__() |
| 40 | |
| 41 | image_HW = make_2tuple(img_size) |
| 42 | patch_HW = make_2tuple(patch_size) |
| 43 | patch_grid_size = ( |
| 44 | image_HW[0] // patch_HW[0], |
| 45 | image_HW[1] // patch_HW[1], |
| 46 | ) |
| 47 | |
| 48 | self.img_size = image_HW |
| 49 | self.patch_size = patch_HW |
| 50 | self.patches_resolution = patch_grid_size |
| 51 | self.num_patches = patch_grid_size[0] * patch_grid_size[1] |
| 52 | |
| 53 | self.in_chans = in_chans |
| 54 | self.embed_dim = embed_dim |
| 55 | |
| 56 | self.flatten_embedding = flatten_embedding |
| 57 | |
| 58 | self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_HW, stride=patch_HW) |
| 59 | self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity() |
| 60 | |
| 61 | def forward(self, x: Tensor) -> Tensor: |
| 62 | _, _, H, W = x.shape |