Fix Automatic Download of Pretrained Weights in DETR (#17712)

* added use_backbone_pretrained

* style fixes

* update

* Update detr.mdx

* Update detr.mdx

* Update detr.mdx

* update using doc py

* Update detr.mdx

* Update src/transformers/models/detr/configuration_detr.py

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

docs/source/en/model_doc/detr.mdx

@@ -113,6 +113,28 @@ Tips:
 - The size of the images will determine the amount of memory being used, and will thus determine the `batch_size`.
   It is advised to use a batch size of 2 per GPU. See [this Github thread](https://github.com/facebookresearch/detr/issues/150) for more info.
 
+There are three ways to instantiate a DETR model (depending on what you prefer):
+  
+Option 1: Instantiate DETR with pre-trained weights for entire model
+```py
+>>> from transformers import DetrForObjectDetection
+
+>>> model = DetrForObjectDetection.from_pretrained("facebook/resnet-50")
+```
+
+Option 2: Instantiate DETR with randomly initialized weights for Transformer, but pre-trained weights for backbone
+```py
+>>> from transformers import DetrConfig, DetrForObjectDetection
+
+>>> config = DetrConfig()
+>>> model = DetrForObjectDetection(config)
+```
+Option 3: Instantiate DETR with randomly initialized weights for backbone + Transformer
+```py
+>>> config = DetrConfig(use_pretrained_backbone=False)
+>>> model = DetrForObjectDetection(config)
+```
+
 As a summary, consider the following table:
 
 | Task | Object detection | Instance segmentation | Panoptic segmentation |
@@ -166,4 +188,4 @@ mean Average Precision (mAP) and Panoptic Quality (PQ). The latter objects are i
 ## DetrForSegmentation
 
 [[autodoc]] DetrForSegmentation
-    - forward
+    - forward