Big file_utils cleanup (#16396)

* Big file_utils cleanup

* This one still needs to be treated separately

docs/source/main_classes/model.mdx

@@ -88,4 +88,4 @@ Due to Pytorch design, this functionality is only available for floating dtypes.
 
 ## Pushing to the Hub
 
-[[autodoc]] file_utils.PushToHubMixin
+[[autodoc]] utils.PushToHubMixin

docs/source/main_classes/output.mdx

@@ -12,7 +12,7 @@ specific language governing permissions and limitations under the License.
 
 # Model outputs
 
-All models have outputs that are instances of subclasses of [`~file_utils.ModelOutput`]. Those are
+All models have outputs that are instances of subclasses of [`~utils.ModelOutput`]. Those are
 data structures containing all the information returned by the model, but that can also be used as tuples or
 dictionaries.
 
@@ -57,7 +57,7 @@ documented on their corresponding model page.
 
 ## ModelOutput
 
-[[autodoc]] file_utils.ModelOutput
+[[autodoc]] utils.ModelOutput
     - to_tuple
 
 ## BaseModelOutput

docs/source/main_classes/trainer.mdx

@@ -40,7 +40,7 @@ The [`Trainer`] contains the basic training loop which supports the above featur
 The [`Trainer`] class is optimized for 🤗 Transformers models and can have surprising behaviors
 when you use it on other models. When using it on your own model, make sure:
 
-- your model always return tuples or subclasses of [`~file_utils.ModelOutput`].
+- your model always return tuples or subclasses of [`~utils.ModelOutput`].
 - your model can compute the loss if a `labels` argument is provided and that loss is returned as the first
   element of the tuple (if your model returns tuples)
 - your model can accept multiple label arguments (use the `label_names` in your [`TrainingArguments`] to indicate their name to the [`Trainer`]) but none of them should be named `"label"`.