MBart: Fix docs and doctests (#22422)

Fix docs and doctests

src/transformers/models/mbart/modeling_mbart.py

@@ -561,7 +561,7 @@ MBART_GENERATION_EXAMPLE = r"""
     >>> inputs = tokenizer(example_english_phrase, return_tensors="pt")
 
     >>> # Translate
-    >>> generated_ids = model.generate(inputs["input_ids"], num_beams=4, max_length=5)
+    >>> generated_ids = model.generate(**inputs, num_beams=4, max_length=5)
     >>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
     '42 este răspuns'
     ```
@@ -1266,7 +1266,8 @@ class MBartModel(MBartPreTrainedModel):
 
 
 @add_start_docstrings(
-    "The MBART Model with a language modeling head. Can be used for summarization.", MBART_START_DOCSTRING
+    "The MBART Model with a language modeling head. Can be used for summarization, after fine-tuning the pretrained models.",
+    MBART_START_DOCSTRING,
 )
 class MBartForConditionalGeneration(MBartPreTrainedModel):
     base_model_prefix = "model"

src/transformers/models/mbart/modeling_tf_mbart.py

@@ -619,37 +619,44 @@ MBART_INPUTS_DOCSTRING = r"""
 """
 
 MBART_GENERATION_EXAMPLE = r"""
-    Summarization example:
+    Translation example:
 
     ```python
-    >>> from transformers import AutoTokenizer, TFMBartForConditionalGeneration, MBartConfig
+    >>> from transformers import AutoTokenizer, TFMBartForConditionalGeneration
 
-    >>> model = TFMBartForConditionalGeneration.from_pretrained("facebook/mbart-large-cc25")
-    >>> tokenizer = AutoTokenizer.from_pretrained("facebook/mbart-large-cc25")
+    >>> model = TFMBartForConditionalGeneration.from_pretrained("facebook/mbart-large-en-ro")
+    >>> tokenizer = AutoTokenizer.from_pretrained("facebook/mbart-large-en-ro")
 
-    >>> ARTICLE_TO_SUMMARIZE = "Meine Freunde sind cool, aber sie essen zu viel Kuchen."
-    >>> inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors="tf")
+    >>> example_english_phrase = "42 is the answer"
+    >>> inputs = tokenizer(example_english_phrase, return_tensors="tf")
 
-    >>> # Generate Summary
-    >>> summary_ids = model.generate(inputs["input_ids"], num_beams=4, max_length=5)
-    >>> print(tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False))
+    >>> # Translate
+    >>> generated_ids = model.generate(**inputs, num_beams=4, max_length=5)
+    >>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+    '42 este răspuns'
     ```
 
     Mask filling example:
 
     ```python
     >>> from transformers import AutoTokenizer, TFMBartForConditionalGeneration
+    >>> import tensorflow as tf
 
-    >>> model = MBartForConditionalGeneration.from_pretrained("facebook/mbart-large-cc25")
+    >>> model = TFMBartForConditionalGeneration.from_pretrained("facebook/mbart-large-cc25")
     >>> tokenizer = AutoTokenizer.from_pretrained("facebook/mbart-large-cc25")
 
     >>> # de_DE is the language symbol id <LID> for German
     >>> TXT = "</s> Meine Freunde sind <mask> nett aber sie essen zu viel Kuchen. </s> de_DE"
-    >>> input_ids = tokenizer([TXT], add_special_tokens=False, return_tensors="tf")["input_ids"]
 
+    >>> input_ids = tokenizer([TXT], add_special_tokens=False, return_tensors="tf")["input_ids"]
     >>> logits = model(input_ids).logits
-    >>> probs = tf.nn.softmax(logits[0])
-    >>> # probs[5] is associated with the mask token
+
+    >>> masked_index = tf.where(input_ids[0] == tokenizer.mask_token_id)[0, 0]
+    >>> probs = tf.nn.softmax(logits[0, masked_index], axis=0)
+    >>> values, predictions = tf.math.top_k(probs, 5)
+
+    >>> tokenizer.decode(predictions).split()
+    ['nett', 'sehr', 'ganz', 'nicht', 'so']
     ```
 """
 
@@ -1299,7 +1306,7 @@ class BiasLayer(tf.keras.layers.Layer):
 
 
 @add_start_docstrings(
-    "The MBART Model with a language modeling head. Can be used for summarization.",
+    "The MBART Model with a language modeling head. Can be used for summarization, after fine-tuning the pretrained models.",
     MBART_START_DOCSTRING,
 )
 class TFMBartForConditionalGeneration(TFMBartPreTrainedModel, TFCausalLanguageModelingLoss):

utils/documentation_tests.txt

@@ -120,6 +120,7 @@ src/transformers/models/maskformer/configuration_maskformer.py
 src/transformers/models/maskformer/modeling_maskformer.py
 src/transformers/models/mbart/configuration_mbart.py
 src/transformers/models/mbart/modeling_mbart.py
+src/transformers/models/mbart/modeling_tf_mbart.py
 src/transformers/models/mctct/configuration_mctct.py
 src/transformers/models/megatron_bert/configuration_megatron_bert.py
 src/transformers/models/mobilebert/configuration_mobilebert.py
@@ -479,4 +480,4 @@ src/transformers/models/m2m_100/tokenization_m2m_100.py
 src/transformers/models/marian/tokenization_marian.py
 src/transformers/models/roformer/tokenization_roformer.py
 src/transformers/models/roformer/tokenization_roformer_fast.py
-src/transformers/models/transfo_xl/tokenization_transfo_xl.py
+src/transformers/models/transfo_xl/tokenization_transfo_xl.py