Add "Leveraging Pretrained Checkpoints for Generation" Seq2Seq models. (#6594)

* add conversion script

* improve conversion script

* make style

* add tryout files

* fix

* update

* add causal bert

* better names

* add tokenizer file as well

* finish causal_bert

* fix small bugs

* improve generate

* change naming

* renaming

* renaming

* renaming

* remove leftover files

* clean files

* add fix tokenizer

* finalize

* correct slow test

* update docs

* small fixes

* fix link

* adapt check repo

* apply sams and sylvains recommendations

* fix import

* implement Lysandres recommendations

* fix logger warn

utils/check_repo.py

@@ -47,6 +47,7 @@ MODEL_NAME_TO_DOC_FILE = {
     "openai": "gpt.rst",
     "transfo_xl": "transformerxl.rst",
     "xlm_roberta": "xlmroberta.rst",
+    "bert_generation": "bertgeneration.rst",
 }
 
 # This is to make sure the transformers module imported is the one in the repo.
@@ -230,6 +231,9 @@ def _get_model_name(module):
     # Secial case for xlm_roberta
     if splits[-1] == "roberta" and splits[-2] == "xlm":
         return "_".join(splits[-2:])
+    # Special case for bert_generation
+    if splits[-1] == "generation" and splits[-2] == "bert":
+        return "_".join(splits[-2:])
     return splits[-1]