[docs] Update to scripts building index.md (#26546)

* build the table in index.md with links to the model_doc

* removed list generation on index.md

* fixed missing models

* make style

utils/check_copies.py

@@ -15,7 +15,7 @@
 """
 Utility that checks whether the copies defined in the library match the original or not. This includes:
 - All code commented with `# Copied from` comments,
-- The list of models in the main README.md matches the ones in the localized READMEs and in the index.md,
+- The list of models in the main README.md matches the ones in the localized READMEs,
 - Files that are registered as full copies of one another in the `FULL_COPIES` constant of this script.
 
 This also checks the list of models in the README is complete (has all models) and add a line to complete if there is
@@ -350,7 +350,7 @@ def is_copy_consistent(filename: str, overwrite: bool = False) -> Optional[List[
 def check_copies(overwrite: bool = False):
     """
     Check every file is copy-consistent with the original. Also check the model list in the main README and other
-    READMEs/index.md are consistent.
+    READMEs are consistent.
 
     Args:
         overwrite (`bool`, *optional*, defaults to `False`):
@@ -517,21 +517,6 @@ def convert_to_localized_md(model_list: str, localized_model_list: str, format_s
     return readmes_match, "\n".join((x[1] for x in sorted_index)) + "\n"
 
 
-def convert_readme_to_index(model_list: str) -> str:
-    """
-    Converts the model list of the README to the index.md format (adapting links to the doc to relative links).
-
-    Args:
-        model_list (`str`): The model list of the main README.
-
-    Returns:
-        `str`: The model list in the format for the index.
-    """
-    # We need to replce both link to the main doc and stable doc (the order of the next two instructions is important).
-    model_list = model_list.replace("https://huggingface.co/docs/transformers/main/", "")
-    return model_list.replace("https://huggingface.co/docs/transformers/", "")
-
-
 def _find_text_in_file(filename: str, start_prompt: str, end_prompt: str) -> Tuple[str, int, int, List[str]]:
     """
     Find the text in a file between two prompts.
@@ -591,19 +576,13 @@ def check_model_list_copy(overwrite: bool = False):
                 "automatically fix them."
             )
 
-    # If the introduction or the conclusion of the list change, the prompts may need to be updated.
-    index_list, start_index, end_index, lines = _find_text_in_file(
-        filename=os.path.join(PATH_TO_DOCS, "index.md"),
-        start_prompt="<!--This list is updated automatically from the README",
-        end_prompt="### Supported frameworks",
-    )
     md_list = get_model_list(
         filename="README.md",
         start_prompt=LOCALIZED_READMES["README.md"]["start_prompt"],
         end_prompt=LOCALIZED_READMES["README.md"]["end_prompt"],
     )
 
-    # Buld the converted Markdown.
+    # Build the converted Markdown.
     converted_md_lists = []
     for filename, value in LOCALIZED_READMES.items():
         _start_prompt = value["start_prompt"]
@@ -615,18 +594,6 @@ def check_model_list_copy(overwrite: bool = False):
 
         converted_md_lists.append((filename, readmes_match, converted_md_list, _start_prompt, _end_prompt))
 
-    # Build the converted index and compare it.
-    converted_md_list = convert_readme_to_index(md_list)
-    if converted_md_list != index_list:
-        if overwrite:
-            with open(os.path.join(PATH_TO_DOCS, "index.md"), "w", encoding="utf-8", newline="\n") as f:
-                f.writelines(lines[:start_index] + [converted_md_list] + lines[end_index:])
-        else:
-            raise ValueError(
-                "The model list in the README changed and the list in `index.md` has not been updated. Run "
-                "`make fix-copies` to fix this."
-            )
-
     # Compare the converted Markdowns
     for converted_md_list in converted_md_lists:
         filename, readmes_match, converted_md, _start_prompt, _end_prompt = converted_md_list

utils/check_table.py

@@ -132,6 +132,46 @@ def _center_text(text: str, width: int) -> str:
     return " " * left_indent + text + " " * right_indent
 
 
+SPECIAL_MODEL_NAME_LINK_MAPPING = {
+    "Data2VecAudio": "[Data2VecAudio](model_doc/data2vec)",
+    "Data2VecText": "[Data2VecText](model_doc/data2vec)",
+    "Data2VecVision": "[Data2VecVision](model_doc/data2vec)",
+    "DonutSwin": "[DonutSwin](model_doc/donut)",
+}
+
+MODEL_NAMES_WITH_SAME_CONFIG = {
+    "BARThez": "BART",
+    "BARTpho": "BART",
+    "BertJapanese": "BERT",
+    "BERTweet": "BERT",
+    "BORT": "BERT",
+    "ByT5": "T5",
+    "CPM": "OpenAI GPT-2",
+    "DePlot": "Pix2Struct",
+    "DialoGPT": "OpenAI GPT-2",
+    "DiT": "BEiT",
+    "FLAN-T5": "T5",
+    "FLAN-UL2": "T5",
+    "HerBERT": "BERT",
+    "LayoutXLM": "LayoutLMv2",
+    "Llama2": "LLaMA",
+    "MatCha": "Pix2Struct",
+    "mBART-50": "mBART",
+    "Megatron-GPT2": "OpenAI GPT-2",
+    "mLUKE": "LUKE",
+    "MMS": "Wav2Vec2",
+    "NLLB": "M2M100",
+    "PhoBERT": "BERT",
+    "T5v1.1": "T5",
+    "TAPEX": "BART",
+    "UL2": "T5",
+    "Wav2Vec2Phoneme": "Wav2Vec2",
+    "XLM-V": "XLM-RoBERTa",
+    "XLS-R": "Wav2Vec2",
+    "XLSR-Wav2Vec2": "Wav2Vec2",
+}
+
+
 def get_model_table_from_auto_modules() -> str:
     """
     Generates an up-to-date model table from the content of the auto modules.
@@ -172,17 +212,27 @@ def get_model_table_from_auto_modules() -> str:
                 attr_name = "".join(camel_case_split(attr_name)[:-1])
 
     # Let's build that table!
-    model_names = list(model_name_to_config.keys())
+    model_names = list(model_name_to_config.keys()) + list(MODEL_NAMES_WITH_SAME_CONFIG.keys())
+
+    # model name to doc link mapping
+    model_names_mapping = transformers_module.models.auto.configuration_auto.MODEL_NAMES_MAPPING
+    model_name_to_link_mapping = {value: f"[{value}](model_doc/{key})" for key, value in model_names_mapping.items()}
+    # update mapping with special model names
+    model_name_to_link_mapping = {
+        k: SPECIAL_MODEL_NAME_LINK_MAPPING[k] if k in SPECIAL_MODEL_NAME_LINK_MAPPING else v
+        for k, v in model_name_to_link_mapping.items()
+    }
 
     # MaskFormerSwin and TimmBackbone are backbones and so not meant to be loaded and used on their own. Instead, they define architectures which can be loaded using the AutoBackbone API.
-    names_to_exclude = ["MaskFormerSwin", "TimmBackbone"]
+    names_to_exclude = ["MaskFormerSwin", "TimmBackbone", "Speech2Text2"]
     model_names = [name for name in model_names if name not in names_to_exclude]
     model_names.sort(key=str.lower)
 
     columns = ["Model", "PyTorch support", "TensorFlow support", "Flax Support"]
     # We'll need widths to properly display everything in the center (+2 is to leave one extra space on each side).
+
     widths = [len(c) + 2 for c in columns]
-    widths[0] = max([len(name) for name in model_names]) + 2
+    widths[0] = max([len(doc_link) for doc_link in model_name_to_link_mapping.values()]) + 2
 
     # Build the table per se
     table = "|" + "|".join([_center_text(c, w) for c, w in zip(columns, widths)]) + "|\n"
@@ -190,10 +240,14 @@ def get_model_table_from_auto_modules() -> str:
     table += "|" + "|".join([":" + "-" * (w - 2) + ":" for w in widths]) + "|\n"
 
     check = {True: "✅", False: "❌"}
+
     for name in model_names:
+        if name in MODEL_NAMES_WITH_SAME_CONFIG.keys():
+            prefix = model_name_to_prefix[MODEL_NAMES_WITH_SAME_CONFIG[name]]
+        else:
             prefix = model_name_to_prefix[name]
         line = [
-            name,
+            model_name_to_link_mapping[name],
             check[pt_models[prefix]],
             check[tf_models[prefix]],
             check[flax_models[prefix]],