Enable dynamic resolution input for Swin Transformer and variants (#30656)

* add interpolation of positional encoding support to swin

* add style changes

* use default image processor and make size a dictionary

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* remove logits testing

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Refactor image size validation logic when interpolation is disabled

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* remove asserts in modeling

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* add dynamic resolution input support to swinv2

* change size to ensure interpolation encoding path is triggered

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* set interpolate_pos_encoding default value to False

* add dynamic resolution input to donut swin

* add dynamic resolution input to maskformer swin

---------

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

src/transformers/models/donut/modeling_donut_swin.py

@@ -166,10 +166,48 @@ class DonutSwinEmbeddings(nn.Module):
         self.norm = nn.LayerNorm(config.embed_dim)
         self.dropout = nn.Dropout(config.hidden_dropout_prob)
 
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
+        """
+        This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher
+        resolution images.
+
+        Source:
+        https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174
+        """
+
+        num_patches = embeddings.shape[1] - 1
+        num_positions = self.position_embeddings.shape[1] - 1
+        if num_patches == num_positions and height == width:
+            return self.position_embeddings
+        class_pos_embed = self.position_embeddings[:, 0]
+        patch_pos_embed = self.position_embeddings[:, 1:]
+        dim = embeddings.shape[-1]
+        h0 = height // self.config.patch_size
+        w0 = width // self.config.patch_size
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        h0, w0 = h0 + 0.1, w0 + 0.1
+        patch_pos_embed = patch_pos_embed.reshape(1, int(math.sqrt(num_positions)), int(math.sqrt(num_positions)), dim)
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            scale_factor=(h0 / math.sqrt(num_positions), w0 / math.sqrt(num_positions)),
+            mode="bicubic",
+            align_corners=False,
+        )
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1)
+
     def forward(
-        self, pixel_values: Optional[torch.FloatTensor], bool_masked_pos: Optional[torch.BoolTensor] = None
+        self,
+        pixel_values: Optional[torch.FloatTensor],
+        bool_masked_pos: Optional[torch.BoolTensor] = None,
+        interpolate_pos_encoding: bool = False,
     ) -> Tuple[torch.Tensor]:
-        embeddings, output_dimensions = self.patch_embeddings(pixel_values)
+        _, num_channels, height, width = pixel_values.shape
+        embeddings, output_dimensions = self.patch_embeddings(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding
+        )
         embeddings = self.norm(embeddings)
         batch_size, seq_len, _ = embeddings.size()
 
@@ -180,6 +218,9 @@ class DonutSwinEmbeddings(nn.Module):
             embeddings = embeddings * (1.0 - mask) + mask_tokens * mask
 
         if self.position_embeddings is not None:
+            if interpolate_pos_encoding:
+                embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+            else:
                 embeddings = embeddings + self.position_embeddings
 
         embeddings = self.dropout(embeddings)
@@ -219,7 +260,9 @@ class DonutSwinPatchEmbeddings(nn.Module):
             pixel_values = nn.functional.pad(pixel_values, pad_values)
         return pixel_values
 
-    def forward(self, pixel_values: Optional[torch.FloatTensor]) -> Tuple[torch.Tensor, Tuple[int]]:
+    def forward(
+        self, pixel_values: Optional[torch.FloatTensor], interpolate_pos_encoding: bool = False
+    ) -> Tuple[torch.Tensor, Tuple[int]]:
         _, num_channels, height, width = pixel_values.shape
         if num_channels != self.num_channels:
             raise ValueError(
@@ -227,6 +270,11 @@ class DonutSwinPatchEmbeddings(nn.Module):
             )
         # pad the input to be divisible by self.patch_size, if needed
         pixel_values = self.maybe_pad(pixel_values, height, width)
+        if not interpolate_pos_encoding and (height != self.image_size[0] or width != self.image_size[1]):
+            raise ValueError(
+                f"Input image size ({height}*{width}) doesn't match model"
+                f" ({self.image_size[0]}*{self.image_size[1]})."
+            )
         embeddings = self.projection(pixel_values)
         _, _, height, width = embeddings.shape
         output_dimensions = (height, width)
@@ -849,6 +897,8 @@ SWIN_INPUTS_DOCSTRING = r"""
         output_hidden_states (`bool`, *optional*):
             Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
             more detail.
+        interpolate_pos_encoding (`bool`, *optional*, defaults to `False`):
+            Whether to interpolate the pre-trained position encodings.
         return_dict (`bool`, *optional*):
             Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
 """
@@ -899,6 +949,7 @@ class DonutSwinModel(DonutSwinPreTrainedModel):
         head_mask: Optional[torch.FloatTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, DonutSwinModelOutput]:
         r"""
@@ -921,7 +972,9 @@ class DonutSwinModel(DonutSwinPreTrainedModel):
         # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
         head_mask = self.get_head_mask(head_mask, len(self.config.depths))
 
-        embedding_output, input_dimensions = self.embeddings(pixel_values, bool_masked_pos=bool_masked_pos)
+        embedding_output, input_dimensions = self.embeddings(
+            pixel_values, bool_masked_pos=bool_masked_pos, interpolate_pos_encoding=interpolate_pos_encoding
+        )
 
         encoder_outputs = self.encoder(
             embedding_output,

src/transformers/models/maskformer/modeling_maskformer_swin.py

@@ -163,11 +163,49 @@ class MaskFormerSwinEmbeddings(nn.Module):
         self.norm = nn.LayerNorm(config.embed_dim)
         self.dropout = nn.Dropout(config.hidden_dropout_prob)
 
-    def forward(self, pixel_values):
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
-        embeddings, output_dimensions = self.patch_embeddings(pixel_values)
+        """
+        This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher
+        resolution images.
+
+        Source:
+        https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174
+        """
+
+        num_patches = embeddings.shape[1] - 1
+        num_positions = self.position_embeddings.shape[1] - 1
+        if num_patches == num_positions and height == width:
+            return self.position_embeddings
+        class_pos_embed = self.position_embeddings[:, 0]
+        patch_pos_embed = self.position_embeddings[:, 1:]
+        dim = embeddings.shape[-1]
+        h0 = height // self.config.patch_size
+        w0 = width // self.config.patch_size
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        h0, w0 = h0 + 0.1, w0 + 0.1
+        patch_pos_embed = patch_pos_embed.reshape(1, int(math.sqrt(num_positions)), int(math.sqrt(num_positions)), dim)
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            scale_factor=(h0 / math.sqrt(num_positions), w0 / math.sqrt(num_positions)),
+            mode="bicubic",
+            align_corners=False,
+        )
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1)
+
+    def forward(self, pixel_values, interpolate_pos_encoding):
+        _, num_channels, height, width = pixel_values.shape
+        embeddings, output_dimensions = self.patch_embeddings(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding
+        )
         embeddings = self.norm(embeddings)
 
         if self.position_embeddings is not None:
+            if interpolate_pos_encoding:
+                embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+            else:
                 embeddings = embeddings + self.position_embeddings
 
         embeddings = self.dropout(embeddings)
@@ -207,7 +245,9 @@ class MaskFormerSwinPatchEmbeddings(nn.Module):
             pixel_values = nn.functional.pad(pixel_values, pad_values)
         return pixel_values
 
-    def forward(self, pixel_values: Optional[torch.FloatTensor]) -> Tuple[torch.Tensor, Tuple[int]]:
+    def forward(
+        self, pixel_values: Optional[torch.FloatTensor], interpolate_pos_encoding: bool = False
+    ) -> Tuple[torch.Tensor, Tuple[int]]:
         _, num_channels, height, width = pixel_values.shape
         if num_channels != self.num_channels:
             raise ValueError(
@@ -215,6 +255,11 @@ class MaskFormerSwinPatchEmbeddings(nn.Module):
             )
         # pad the input to be divisible by self.patch_size, if needed
         pixel_values = self.maybe_pad(pixel_values, height, width)
+        if not interpolate_pos_encoding and (height != self.image_size[0] or width != self.image_size[1]):
+            raise ValueError(
+                f"Input image size ({height}*{width}) doesn't match model"
+                f" ({self.image_size[0]}*{self.image_size[1]})."
+            )
         embeddings = self.projection(pixel_values)
         _, _, height, width = embeddings.shape
         output_dimensions = (height, width)
@@ -780,6 +825,7 @@ class MaskFormerSwinModel(MaskFormerSwinPreTrainedModel):
         head_mask=None,
         output_attentions=None,
         output_hidden_states=None,
+        interpolate_pos_encoding=False,
         return_dict=None,
     ):
         output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
@@ -798,7 +844,9 @@ class MaskFormerSwinModel(MaskFormerSwinPreTrainedModel):
         # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
         head_mask = self.get_head_mask(head_mask, len(self.config.depths))
 
-        embedding_output, input_dimensions = self.embeddings(pixel_values)
+        embedding_output, input_dimensions = self.embeddings(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding
+        )
 
         encoder_outputs = self.encoder(
             embedding_output,

src/transformers/models/swin/modeling_swin.py

@@ -252,10 +252,48 @@ class SwinEmbeddings(nn.Module):
         self.norm = nn.LayerNorm(config.embed_dim)
         self.dropout = nn.Dropout(config.hidden_dropout_prob)
 
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
+        """
+        This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher
+        resolution images.
+
+        Source:
+        https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174
+        """
+
+        num_patches = embeddings.shape[1] - 1
+        num_positions = self.position_embeddings.shape[1] - 1
+        if num_patches == num_positions and height == width:
+            return self.position_embeddings
+        class_pos_embed = self.position_embeddings[:, 0]
+        patch_pos_embed = self.position_embeddings[:, 1:]
+        dim = embeddings.shape[-1]
+        h0 = height // self.config.patch_size
+        w0 = width // self.config.patch_size
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        h0, w0 = h0 + 0.1, w0 + 0.1
+        patch_pos_embed = patch_pos_embed.reshape(1, int(math.sqrt(num_positions)), int(math.sqrt(num_positions)), dim)
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            scale_factor=(h0 / math.sqrt(num_positions), w0 / math.sqrt(num_positions)),
+            mode="bicubic",
+            align_corners=False,
+        )
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1)
+
     def forward(
-        self, pixel_values: Optional[torch.FloatTensor], bool_masked_pos: Optional[torch.BoolTensor] = None
+        self,
+        pixel_values: Optional[torch.FloatTensor],
+        bool_masked_pos: Optional[torch.BoolTensor] = None,
+        interpolate_pos_encoding: bool = False,
     ) -> Tuple[torch.Tensor]:
-        embeddings, output_dimensions = self.patch_embeddings(pixel_values)
+        _, num_channels, height, width = pixel_values.shape
+        embeddings, output_dimensions = self.patch_embeddings(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding
+        )
         embeddings = self.norm(embeddings)
         batch_size, seq_len, _ = embeddings.size()
 
@@ -266,6 +304,9 @@ class SwinEmbeddings(nn.Module):
             embeddings = embeddings * (1.0 - mask) + mask_tokens * mask
 
         if self.position_embeddings is not None:
+            if interpolate_pos_encoding:
+                embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+            else:
                 embeddings = embeddings + self.position_embeddings
 
         embeddings = self.dropout(embeddings)
@@ -304,7 +345,9 @@ class SwinPatchEmbeddings(nn.Module):
             pixel_values = nn.functional.pad(pixel_values, pad_values)
         return pixel_values
 
-    def forward(self, pixel_values: Optional[torch.FloatTensor]) -> Tuple[torch.Tensor, Tuple[int]]:
+    def forward(
+        self, pixel_values: Optional[torch.FloatTensor], interpolate_pos_encoding: bool = False
+    ) -> Tuple[torch.Tensor, Tuple[int]]:
         _, num_channels, height, width = pixel_values.shape
         if num_channels != self.num_channels:
             raise ValueError(
@@ -312,6 +355,11 @@ class SwinPatchEmbeddings(nn.Module):
             )
         # pad the input to be divisible by self.patch_size, if needed
         pixel_values = self.maybe_pad(pixel_values, height, width)
+        if not interpolate_pos_encoding and (height != self.image_size[0] or width != self.image_size[1]):
+            raise ValueError(
+                f"Input image size ({height}*{width}) doesn't match model"
+                f" ({self.image_size[0]}*{self.image_size[1]})."
+            )
         embeddings = self.projection(pixel_values)
         _, _, height, width = embeddings.shape
         output_dimensions = (height, width)
@@ -924,6 +972,8 @@ SWIN_INPUTS_DOCSTRING = r"""
         output_hidden_states (`bool`, *optional*):
             Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
             more detail.
+        interpolate_pos_encoding (`bool`, *optional*, defaults to `False`):
+            Whether to interpolate the pre-trained position encodings.
         return_dict (`bool`, *optional*):
             Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
 """
@@ -981,6 +1031,7 @@ class SwinModel(SwinPreTrainedModel):
         head_mask: Optional[torch.FloatTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, SwinModelOutput]:
         r"""
@@ -1003,7 +1054,9 @@ class SwinModel(SwinPreTrainedModel):
         # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
         head_mask = self.get_head_mask(head_mask, len(self.config.depths))
 
-        embedding_output, input_dimensions = self.embeddings(pixel_values, bool_masked_pos=bool_masked_pos)
+        embedding_output, input_dimensions = self.embeddings(
+            pixel_values, bool_masked_pos=bool_masked_pos, interpolate_pos_encoding=interpolate_pos_encoding
+        )
 
         encoder_outputs = self.encoder(
             embedding_output,
@@ -1074,6 +1127,7 @@ class SwinForMaskedImageModeling(SwinPreTrainedModel):
         head_mask: Optional[torch.FloatTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, SwinMaskedImageModelingOutput]:
         r"""
@@ -1113,6 +1167,7 @@ class SwinForMaskedImageModeling(SwinPreTrainedModel):
             head_mask=head_mask,
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
+            interpolate_pos_encoding=interpolate_pos_encoding,
             return_dict=return_dict,
         )
 
@@ -1156,6 +1211,14 @@ class SwinForMaskedImageModeling(SwinPreTrainedModel):
     """
     Swin Model transformer with an image classification head on top (a linear layer on top of the final hidden state of
     the [CLS] token) e.g. for ImageNet.
+
+    <Tip>
+
+        Note that it's possible to fine-tune Swin on higher resolution images than the ones it has been trained on, by
+        setting `interpolate_pos_encoding` to `True` in the forward of the model. This will interpolate the pre-trained
+        position embeddings to the higher resolution.
+
+    </Tip>
     """,
     SWIN_START_DOCSTRING,
 )
@@ -1188,6 +1251,7 @@ class SwinForImageClassification(SwinPreTrainedModel):
         labels: Optional[torch.LongTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, SwinImageClassifierOutput]:
         r"""
@@ -1203,6 +1267,7 @@ class SwinForImageClassification(SwinPreTrainedModel):
             head_mask=head_mask,
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
+            interpolate_pos_encoding=interpolate_pos_encoding,
             return_dict=return_dict,
         )
 

src/transformers/models/swinv2/modeling_swinv2.py

@@ -295,10 +295,48 @@ class Swinv2Embeddings(nn.Module):
         self.norm = nn.LayerNorm(config.embed_dim)
         self.dropout = nn.Dropout(config.hidden_dropout_prob)
 
+    def interpolate_pos_encoding(self, embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor:
+        """
+        This method allows to interpolate the pre-trained position encodings, to be able to use the model on higher
+        resolution images.
+
+        Source:
+        https://github.com/facebookresearch/dino/blob/de9ee3df6cf39fac952ab558447af1fa1365362a/vision_transformer.py#L174
+        """
+
+        num_patches = embeddings.shape[1] - 1
+        num_positions = self.position_embeddings.shape[1] - 1
+        if num_patches == num_positions and height == width:
+            return self.position_embeddings
+        class_pos_embed = self.position_embeddings[:, 0]
+        patch_pos_embed = self.position_embeddings[:, 1:]
+        dim = embeddings.shape[-1]
+        h0 = height // self.config.patch_size
+        w0 = width // self.config.patch_size
+        # we add a small number to avoid floating point error in the interpolation
+        # see discussion at https://github.com/facebookresearch/dino/issues/8
+        h0, w0 = h0 + 0.1, w0 + 0.1
+        patch_pos_embed = patch_pos_embed.reshape(1, int(math.sqrt(num_positions)), int(math.sqrt(num_positions)), dim)
+        patch_pos_embed = patch_pos_embed.permute(0, 3, 1, 2)
+        patch_pos_embed = nn.functional.interpolate(
+            patch_pos_embed,
+            scale_factor=(h0 / math.sqrt(num_positions), w0 / math.sqrt(num_positions)),
+            mode="bicubic",
+            align_corners=False,
+        )
+        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
+        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1)
+
     def forward(
-        self, pixel_values: Optional[torch.FloatTensor], bool_masked_pos: Optional[torch.BoolTensor] = None
+        self,
+        pixel_values: Optional[torch.FloatTensor],
+        bool_masked_pos: Optional[torch.BoolTensor] = None,
+        interpolate_pos_encoding: bool = False,
     ) -> Tuple[torch.Tensor]:
-        embeddings, output_dimensions = self.patch_embeddings(pixel_values)
+        _, num_channels, height, width = pixel_values.shape
+        embeddings, output_dimensions = self.patch_embeddings(
+            pixel_values, interpolate_pos_encoding=interpolate_pos_encoding
+        )
         embeddings = self.norm(embeddings)
         batch_size, seq_len, _ = embeddings.size()
 
@@ -309,6 +347,9 @@ class Swinv2Embeddings(nn.Module):
             embeddings = embeddings * (1.0 - mask) + mask_tokens * mask
 
         if self.position_embeddings is not None:
+            if interpolate_pos_encoding:
+                embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+            else:
                 embeddings = embeddings + self.position_embeddings
 
         embeddings = self.dropout(embeddings)
@@ -348,7 +389,9 @@ class Swinv2PatchEmbeddings(nn.Module):
             pixel_values = nn.functional.pad(pixel_values, pad_values)
         return pixel_values
 
-    def forward(self, pixel_values: Optional[torch.FloatTensor]) -> Tuple[torch.Tensor, Tuple[int]]:
+    def forward(
+        self, pixel_values: Optional[torch.FloatTensor], interpolate_pos_encoding: bool = False
+    ) -> Tuple[torch.Tensor, Tuple[int]]:
         _, num_channels, height, width = pixel_values.shape
         if num_channels != self.num_channels:
             raise ValueError(
@@ -356,6 +399,11 @@ class Swinv2PatchEmbeddings(nn.Module):
             )
         # pad the input to be divisible by self.patch_size, if needed
         pixel_values = self.maybe_pad(pixel_values, height, width)
+        if not interpolate_pos_encoding and (height != self.image_size[0] or width != self.image_size[1]):
+            raise ValueError(
+                f"Input image size ({height}*{width}) doesn't match model"
+                f" ({self.image_size[0]}*{self.image_size[1]})."
+            )
         embeddings = self.projection(pixel_values)
         _, _, height, width = embeddings.shape
         output_dimensions = (height, width)
@@ -979,6 +1027,8 @@ SWINV2_INPUTS_DOCSTRING = r"""
         output_hidden_states (`bool`, *optional*):
             Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
             more detail.
+        interpolate_pos_encoding (`bool`, *optional*, default `False`):
+            Whether to interpolate the pre-trained position encodings.
         return_dict (`bool`, *optional*):
             Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
 """
@@ -1031,6 +1081,7 @@ class Swinv2Model(Swinv2PreTrainedModel):
         head_mask: Optional[torch.FloatTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, Swinv2ModelOutput]:
         r"""
@@ -1053,7 +1104,9 @@ class Swinv2Model(Swinv2PreTrainedModel):
         # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
         head_mask = self.get_head_mask(head_mask, len(self.config.depths))
 
-        embedding_output, input_dimensions = self.embeddings(pixel_values, bool_masked_pos=bool_masked_pos)
+        embedding_output, input_dimensions = self.embeddings(
+            pixel_values, bool_masked_pos=bool_masked_pos, interpolate_pos_encoding=interpolate_pos_encoding
+        )
 
         encoder_outputs = self.encoder(
             embedding_output,
@@ -1126,6 +1179,7 @@ class Swinv2ForMaskedImageModeling(Swinv2PreTrainedModel):
         head_mask: Optional[torch.FloatTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, Swinv2MaskedImageModelingOutput]:
         r"""
@@ -1165,6 +1219,7 @@ class Swinv2ForMaskedImageModeling(Swinv2PreTrainedModel):
             head_mask=head_mask,
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
+            interpolate_pos_encoding=interpolate_pos_encoding,
             return_dict=return_dict,
         )
 
@@ -1208,6 +1263,14 @@ class Swinv2ForMaskedImageModeling(Swinv2PreTrainedModel):
     """
     Swinv2 Model transformer with an image classification head on top (a linear layer on top of the final hidden state
     of the [CLS] token) e.g. for ImageNet.
+
+    <Tip>
+
+        Note that it's possible to fine-tune SwinV2 on higher resolution images than the ones it has been trained on, by
+        setting `interpolate_pos_encoding` to `True` in the forward of the model. This will interpolate the pre-trained
+        position embeddings to the higher resolution.
+
+    </Tip>
     """,
     SWINV2_START_DOCSTRING,
 )
@@ -1241,6 +1304,7 @@ class Swinv2ForImageClassification(Swinv2PreTrainedModel):
         labels: Optional[torch.LongTensor] = None,
         output_attentions: Optional[bool] = None,
         output_hidden_states: Optional[bool] = None,
+        interpolate_pos_encoding: bool = False,
         return_dict: Optional[bool] = None,
     ) -> Union[Tuple, Swinv2ImageClassifierOutput]:
         r"""
@@ -1256,6 +1320,7 @@ class Swinv2ForImageClassification(Swinv2PreTrainedModel):
             head_mask=head_mask,
             output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
+            interpolate_pos_encoding=interpolate_pos_encoding,
             return_dict=return_dict,
         )
 

tests/models/swin/test_modeling_swin.py

@@ -493,6 +493,26 @@ class SwinModelIntegrationTest(unittest.TestCase):
         expected_slice = torch.tensor([-0.0948, -0.6454, -0.0921]).to(torch_device)
         self.assertTrue(torch.allclose(outputs.logits[0, :3], expected_slice, atol=1e-4))
 
+    @slow
+    def test_inference_interpolate_pos_encoding(self):
+        # Swin models have an `interpolate_pos_encoding` argument in their forward method,
+        # allowing to interpolate the pre-trained position embeddings in order to use
+        # the model on higher resolutions.
+        model = SwinModel.from_pretrained("microsoft/swin-tiny-patch4-window7-224").to(torch_device)
+
+        image_processor = self.default_image_processor
+        image = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png")
+        inputs = image_processor(images=image, size={"height": 481, "width": 481}, return_tensors="pt")
+        pixel_values = inputs.pixel_values.to(torch_device)
+
+        # forward pass
+        with torch.no_grad():
+            outputs = model(pixel_values, interpolate_pos_encoding=True)
+
+        # verify the logits
+        expected_shape = torch.Size((1, 256, 768))
+        self.assertEqual(outputs.last_hidden_state.shape, expected_shape)
+
 
 @require_torch
 class SwinBackboneTest(unittest.TestCase, BackboneTesterMixin):

tests/models/swinv2/test_modeling_swinv2.py

@@ -485,6 +485,26 @@ class Swinv2ModelIntegrationTest(unittest.TestCase):
         expected_slice = torch.tensor([-0.3947, -0.4306, 0.0026]).to(torch_device)
         self.assertTrue(torch.allclose(outputs.logits[0, :3], expected_slice, atol=1e-4))
 
+    @slow
+    def test_inference_interpolate_pos_encoding(self):
+        # Swinv2 models have an `interpolate_pos_encoding` argument in their forward method,
+        # allowing to interpolate the pre-trained position embeddings in order to use
+        # the model on higher resolutions.
+        model = Swinv2Model.from_pretrained("microsoft/swinv2-tiny-patch4-window8-256").to(torch_device)
+
+        image_processor = self.default_image_processor
+        image = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png")
+        inputs = image_processor(images=image, size={"height": 481, "width": 481}, return_tensors="pt")
+        pixel_values = inputs.pixel_values.to(torch_device)
+
+        # forward pass
+        with torch.no_grad():
+            outputs = model(pixel_values, interpolate_pos_encoding=True)
+
+        # verify the logits
+        expected_shape = torch.Size((1, 256, 768))
+        self.assertEqual(outputs.last_hidden_state.shape, expected_shape)
+
 
 @require_torch
 class Swinv2BackboneTest(unittest.TestCase, BackboneTesterMixin):