Create dummy models (#19901)

* create dummy models

* quality

* update

* update

* Make Wav2Vec2Conformer work

* style

* deal with models with text_config and vision_config

* apply suggestions

* Composite models

* style

* style

* fix shape issue

* fix shape issue

* For VisionTextDualEncoderModel

* show_progress=False when converting tokenizers

* Fix for OwlViT

* Fix for VisualBert

* Update

* final
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

utils/check_config_docstrings.py

@@ -49,11 +49,8 @@ CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK = {
 }
 
 
-def check_config_docstrings_have_checkpoints():
-    configs_without_checkpoint = []
-
-    for config_class in list(CONFIG_MAPPING.values()):
-        checkpoint_found = False
+def get_checkpoint_from_config_class(config_class):
+    checkpoint = None
 
     # source code of `config_class`
     config_source = inspect.getsource(config_class)
@@ -67,11 +64,20 @@ def check_config_docstrings_have_checkpoints():
         # verify the checkpoint name corresponds to the checkpoint link
         ckpt_link_from_name = f"https://huggingface.co/{ckpt_name}"
         if ckpt_link == ckpt_link_from_name:
-                checkpoint_found = True
+            checkpoint = ckpt_name
             break
 
+    return checkpoint
+
+
+def check_config_docstrings_have_checkpoints():
+    configs_without_checkpoint = []
+
+    for config_class in list(CONFIG_MAPPING.values()):
+        checkpoint = get_checkpoint_from_config_class(config_class)
+
         name = config_class.__name__
-        if not checkpoint_found and name not in CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK:
+        if checkpoint is None and name not in CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK:
             configs_without_checkpoint.append(name)
 
     if len(configs_without_checkpoint) > 0: