Remove a TF usage warning and rework the documentation (huggingface#9756

)

* Rework documentation

* Update the template

* Trigger CI

* Restore the warning but with the TF logger

* Update convbert doc

src/transformers/modeling_tf_utils.py

@@ -44,6 +44,7 @@
 
 
 logger = logging.get_logger(__name__)
+tf_logger = tf.get_logger()
 
 
 class TFModelUtilsMixin:
@@ -285,7 +286,7 @@ def booleans_processing(config, **kwargs):
             or kwargs["output_hidden_states"] is not None
             or ("use_cache" in kwargs and kwargs["use_cache"] is not None)
         ):
-            tf.print(
+            tf_logger.warn(
                 "The parameters `output_attentions`, `output_hidden_states` and `use_cache` cannot be updated when calling a model."
                 "They have to be set to True/False in the config object (i.e.: `config=XConfig.from_pretrained('name', output_attentions=True)`)."
             )
@@ -294,7 +295,7 @@ def booleans_processing(config, **kwargs):
         final_booleans["output_hidden_states"] = config.output_hidden_states
 
         if kwargs["return_dict"] is not None:
-            tf.print("The parameter `return_dict` cannot be set in graph mode and will always be set to `True`.")
+            tf_logger.warn("The parameter `return_dict` cannot be set in graph mode and will always be set to `True`.")
         final_booleans["return_dict"] = True
 
         if "use_cache" in kwargs:

src/transformers/models/albert/modeling_tf_albert.py

@@ -802,12 +802,15 @@ class TFAlbertForPreTrainingOutput(ModelOutput):
             vectors than the model's internal embedding lookup matrix.
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).

src/transformers/models/bart/modeling_tf_bart.py

@@ -574,12 +574,15 @@ def serving(self, inputs):
             decoding (see :obj:`past_key_values`). Set to :obj:`False` during training, :obj:`True` during generation
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).

src/transformers/models/bert/modeling_tf_bert.py

@@ -881,12 +881,15 @@ class TFBertForPreTrainingOutput(ModelOutput):
             vectors than the model's internal embedding lookup matrix.
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).

src/transformers/models/blenderbot/modeling_tf_blenderbot.py

@@ -571,12 +571,15 @@ def serving(self, inputs):
             decoding (see :obj:`past_key_values`). Set to :obj:`False` during training, :obj:`True` during generation
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).
@@ -661,12 +664,18 @@ def call(
                 into associated vectors than the model's internal embedding lookup matrix.
             output_attentions (:obj:`bool`, `optional`):
                 Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under
-                returned tensors for more detail.
+                returned tensors for more detail. This argument can be used only in eager mode, in graph mode the value
+                in the config will be used instead.
             output_hidden_states (:obj:`bool`, `optional`):
                 Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors
-                for more detail.
+                for more detail. This argument can be used only in eager mode, in graph mode the value in the config
+                will be used instead.
             return_dict (:obj:`bool`, `optional`):
-                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+                argument can be used in eager mode, in graph mode the value will always be set to True.
+            training (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to use the model in training mode (some modules like dropout modules have different
+                behaviors between training and evaluation).
         """
         inputs = input_processing(
             func=self.call,
@@ -854,12 +863,18 @@ def call(
                 into associated vectors than the model's internal embedding lookup matrix.
             output_attentions (:obj:`bool`, `optional`):
                 Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under
-                returned tensors for more detail.
+                returned tensors for more detail. This argument can be used only in eager mode, in graph mode the value
+                in the config will be used instead.
             output_hidden_states (:obj:`bool`, `optional`):
                 Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors
-                for more detail.
+                for more detail. This argument can be used only in eager mode, in graph mode the value in the config
+                will be used instead.
             return_dict (:obj:`bool`, `optional`):
-                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+                argument can be used in eager mode, in graph mode the value will always be set to True.
+            training (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to use the model in training mode (some modules like dropout modules have different
+                behaviors between training and evaluation).
         """
         inputs = input_processing(
             func=self.call,

src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py

@@ -576,12 +576,15 @@ def serving(self, inputs):
             decoding (see :obj:`past_key_values`). Set to :obj:`False` during training, :obj:`True` during generation
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).
@@ -666,12 +669,18 @@ def call(
                 into associated vectors than the model's internal embedding lookup matrix.
             output_attentions (:obj:`bool`, `optional`):
                 Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under
-                returned tensors for more detail.
+                returned tensors for more detail. This argument can be used only in eager mode, in graph mode the value
+                in the config will be used instead.
             output_hidden_states (:obj:`bool`, `optional`):
                 Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors
-                for more detail.
+                for more detail. This argument can be used only in eager mode, in graph mode the value in the config
+                will be used instead.
             return_dict (:obj:`bool`, `optional`):
-                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+                argument can be used in eager mode, in graph mode the value will always be set to True.
+            training (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to use the model in training mode (some modules like dropout modules have different
+                behaviors between training and evaluation).
         """
         inputs = input_processing(
             func=self.call,
@@ -859,12 +868,18 @@ def call(
                 into associated vectors than the model's internal embedding lookup matrix.
             output_attentions (:obj:`bool`, `optional`):
                 Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under
-                returned tensors for more detail.
+                returned tensors for more detail. This argument can be used only in eager mode, in graph mode the value
+                in the config will be used instead.
             output_hidden_states (:obj:`bool`, `optional`):
                 Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors
-                for more detail.
+                for more detail. This argument can be used only in eager mode, in graph mode the value in the config
+                will be used instead.
             return_dict (:obj:`bool`, `optional`):
-                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+                Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+                argument can be used in eager mode, in graph mode the value will always be set to True.
+            training (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to use the model in training mode (some modules like dropout modules have different
+                behaviors between training and evaluation).
         """
         inputs = input_processing(
             func=self.call,

src/transformers/models/convbert/modeling_tf_convbert.py

@@ -803,12 +803,15 @@ class TFConvBertPreTrainedModel(TFPreTrainedModel):
             vectors than the model's internal embedding lookup matrix.
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).

src/transformers/models/ctrl/modeling_tf_ctrl.py

@@ -516,12 +516,15 @@ class TFCTRLPreTrainedModel(TFPreTrainedModel):
             ``past``).
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).

src/transformers/models/distilbert/modeling_tf_distilbert.py

@@ -585,12 +585,15 @@ def serving(self, inputs):
             vectors than the model's internal embedding lookup matrix.
         output_attentions (:obj:`bool`, `optional`):
             Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
-            tensors for more detail.
+            tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
+            config will be used instead.
         output_hidden_states (:obj:`bool`, `optional`):
             Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
-            more detail.
+            more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
+            used instead.
         return_dict (:obj:`bool`, `optional`):
-            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. This
+            argument can be used in eager mode, in graph mode the value will always be set to True.
         training (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether or not to use the model in training mode (some modules like dropout modules have different
             behaviors between training and evaluation).