Improve vision models docs (#19103)

* Add tips

* Add BEiT figure

* Fix URL

* Move tip to start

* Add tip to TF model as well
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

docs/source/en/model_doc/beit.mdx

@@ -59,6 +59,11 @@ Tips:
   `use_relative_position_bias` attribute of [`BeitConfig`] to `True` in order to add
   position embeddings.
 
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/beit_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> BEiT pre-training. Taken from the <a href="https://arxiv.org/abs/2106.08254">original paper.</a> </small>
+
 This model was contributed by [nielsr](https://huggingface.co/nielsr). The JAX/FLAX version of this model was
 contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/beit).
 

docs/source/en/model_doc/vit.mdx

@@ -12,13 +12,6 @@ specific language governing permissions and limitations under the License.
 
 # Vision Transformer (ViT)
 
-<Tip>
-
-This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
-breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
-
-</Tip>
-
 ## Overview
 
 The Vision Transformer (ViT) model was proposed in [An Image is Worth 16x16 Words: Transformers for Image Recognition
@@ -63,6 +56,11 @@ Tips:
   language modeling). With this approach, the smaller ViT-B/16 model achieves 79.9% accuracy on ImageNet, a significant
   improvement of 2% to training from scratch, but still 4% behind supervised pre-training.
 
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/vit_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> ViT architecture. Taken from the <a href="https://arxiv.org/abs/2010.11929">original paper.</a> </small>
+
 Following the original Vision Transformer, some follow-up works have been made:
 
 - [DeiT](deit) (Data-efficient Image Transformers) by Facebook AI. DeiT models are distilled vision transformers.

docs/source/en/model_doc/xclip.mdx

@@ -23,7 +23,8 @@ The abstract from the paper is the following:
 
 Tips:
 
-- Usage of X-CLIP is identical to CLIP.
+- Usage of X-CLIP is identical to [CLIP](clip).
+- Demo notebooks for X-CLIP can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/X-CLIP).
 
 <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/xclip_architecture.png"
 alt="drawing" width="600"/> 

src/transformers/models/deit/modeling_deit.py

@@ -555,8 +555,15 @@ class DeiTPooler(nn.Module):
 
 
 @add_start_docstrings(
-    "DeiT Model with a decoder on top for masked image modeling, as proposed in"
-    " [SimMIM](https://arxiv.org/abs/2111.09886).",
+    """DeiT Model with a decoder on top for masked image modeling, as proposed in [SimMIM](https://arxiv.org/abs/2111.09886).
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+    """,
     DEIT_START_DOCSTRING,
 )
 class DeiTForMaskedImageModeling(DeiTPreTrainedModel):

src/transformers/models/swin/modeling_swin.py

@@ -1007,8 +1007,15 @@ class SwinModel(SwinPreTrainedModel):
 
 
 @add_start_docstrings(
-    "Swin Model with a decoder on top for masked image modeling, as proposed in"
-    " [SimMIM](https://arxiv.org/abs/2111.09886).",
+    """Swin Model with a decoder on top for masked image modeling, as proposed in [SimMIM](https://arxiv.org/abs/2111.09886).
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+    """,
     SWIN_START_DOCSTRING,
 )
 class SwinForMaskedImageModeling(SwinPreTrainedModel):

src/transformers/models/swinv2/modeling_swinv2.py

@@ -1087,8 +1087,16 @@ class Swinv2Model(Swinv2PreTrainedModel):
 
 
 @add_start_docstrings(
-    "Swinv2 Model with a decoder on top for masked image modeling, as proposed in"
-    " [SimMIM](https://arxiv.org/abs/2111.09886).",
+    """Swinv2 Model with a decoder on top for masked image modeling, as proposed in
+[SimMIM](https://arxiv.org/abs/2111.09886).
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+    """,
     SWINV2_START_DOCSTRING,
 )
 # Copied from transformers.models.swin.modeling_swin.SwinForMaskedImageModeling with SWIN->SWINV2,Swin->Swinv2,swin->swinv2,224->256,window7->window8

src/transformers/models/vit/modeling_tf_vit.py

@@ -733,6 +733,14 @@ class TFViTPooler(tf.keras.layers.Layer):
     """
     ViT 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 ViT 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>
     """,
     VIT_START_DOCSTRING,
 )

src/transformers/models/vit/modeling_vit.py

@@ -597,8 +597,15 @@ class ViTPooler(nn.Module):
 
 
 @add_start_docstrings(
-    "ViT Model with a decoder on top for masked image modeling, as proposed in"
-    " [SimMIM](https://arxiv.org/abs/2111.09886).",
+    """ViT Model with a decoder on top for masked image modeling, as proposed in [SimMIM](https://arxiv.org/abs/2111.09886).
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+    """,
     VIT_START_DOCSTRING,
 )
 class ViTForMaskedImageModeling(ViTPreTrainedModel):
@@ -712,6 +719,14 @@ class ViTForMaskedImageModeling(ViTPreTrainedModel):
     """
     ViT 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 ViT 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>
     """,
     VIT_START_DOCSTRING,
 )

src/transformers/models/vit_mae/modeling_vit_mae.py

@@ -837,7 +837,16 @@ class ViTMAEDecoder(nn.Module):
 
 
 @add_start_docstrings(
-    "The ViTMAE Model transformer with the decoder on top for self-supervised pre-training.",
+    """The ViTMAE Model transformer with the decoder on top for self-supervised pre-training.
+
+    <Tip>
+
+    Note that we provide a script to pre-train this model on custom data in our [examples
+    directory](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining).
+
+    </Tip>
+
+    """,
     VIT_MAE_START_DOCSTRING,
 )
 class ViTMAEForPreTraining(ViTMAEPreTrainedModel):