Skip to content

TF: Finalize unpack_inputs-related changes #16499

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gante merged 14 commits into from
Apr 4, 2022
Merged

TF: Finalize unpack_inputs-related changes #16499

Show file tree
Hide file tree
Changes from 13 commits
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
558ee00
Add unpack_inputs to remaining models
gante Mar 30, 2022
01c4539
tmp commit
gante Mar 30, 2022
801fd1d
remove breakpoint
gante Mar 30, 2022
821baa7
add unpack_inputs
gante Mar 30, 2022
3e912f7
removed kwargs up to clip
gante Mar 30, 2022
a68901b
removed kwags up to longformer
gante Mar 30, 2022
9bc6e76
remove unused kwargs in remaining models; inputs_processing is now pr…
gante Mar 30, 2022
87b7ab5
forgot to pop empty kwargs
gante Mar 31, 2022
ac1f936
handle models without attention
gante Mar 31, 2022
93c2a65
derp
gante Mar 31, 2022
df4da79
fix t5 tests
gante Apr 1, 2022
364284c
Fix equivalence tests
gante Apr 1, 2022
11e8ca4
Merge branch 'main' into encoder_decoder_unpack_inputs
gante Apr 1, 2022
a8d859d
PR comments
gante Apr 4, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 18 additions & 13 deletions src/transformers/modeling_tf_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -312,9 +312,10 @@ def booleans_processing(config, **kwargs):

if tf.executing_eagerly():
# Pure conv models (such as ConvNext) do not have `output_attentions`
final_booleans["output_attentions"] = kwargs.get("output_attentions", None)
if final_booleans["output_attentions"] is None:
final_booleans["output_attentions"] = config.output_attentions
if "output_attentions" in kwargs:
final_booleans["output_attentions"] = (
kwargs["output_attentions"] if kwargs["output_attentions"] is not None else config.output_attentions
)
Comment on lines +316 to +319
Copy link
Contributor Author

@gante gante Apr 1, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The previous version was passing down final_booleans["output_attentions"]=False in pure conv models, which would set the output_attentions argument to False. The new version results in no argument, which is the desired behavior.

Copy link
Collaborator

@sgugger sgugger Apr 4, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you add a comment that "output_attentions" will be in kwargs, with a value of None if unset? That change made me pause for a couple of minutes.

final_booleans["output_hidden_states"] = (
kwargs["output_hidden_states"]
if kwargs["output_hidden_states"] is not None
Expand All @@ -329,7 +330,9 @@ def booleans_processing(config, **kwargs):
kwargs["use_cache"] if kwargs["use_cache"] is not None else getattr(config, "use_cache", None)
)
else:
final_booleans["output_attentions"] = config.output_attentions
# Pure conv models (such as ConvNext) do not have `output_attentions`
if "output_attentions" in kwargs:
final_booleans["output_attentions"] = config.output_attentions
final_booleans["output_hidden_states"] = config.output_hidden_states

if kwargs.get("return_dict", None) not in (None, True):
Expand Down Expand Up @@ -373,7 +376,7 @@ def run_call_with_unpacked_inputs(self, *args, **kwargs):
# process the inputs and call the wrapped function
main_input_name = getattr(self, "main_input_name", func.__code__.co_varnames[1])
main_input = fn_args_and_kwargs.pop(main_input_name, None)
unpacked_inputs = input_processing(func, self.config, main_input, **fn_args_and_kwargs)
unpacked_inputs = _input_processing(func, self.config, main_input, **fn_args_and_kwargs)
return func(self, **unpacked_inputs)

# Keras enforces the first layer argument to be passed, and checks it through `inspect.getfullargspec()`. This
Expand All @@ -384,7 +387,7 @@ def run_call_with_unpacked_inputs(self, *args, **kwargs):
return run_call_with_unpacked_inputs


def input_processing(func, config, input_ids, **kwargs):
def _input_processing(func, config, input_ids, **kwargs):
"""
Process the input of each TensorFlow model including the booleans. In case of a list of symbolic inputs, each input
has to be named accordingly to the parameters name, i.e. `input_ids = tf.keras.Input(shape=(128,), dtype='int32',
Expand All @@ -402,7 +405,7 @@ def input_processing(func, config, input_ids, **kwargs):
Two lists, one for the missing layers, and another one for the unexpected layers.
"""
signature = dict(inspect.signature(func).parameters)
signature.pop("kwargs", None)
has_kwargs = bool(signature.pop("kwargs", None))
signature.pop("self", None)
parameter_names = list(signature.keys())
output = {}
Expand Down Expand Up @@ -432,12 +435,14 @@ def input_processing(func, config, input_ids, **kwargs):
elif "past_key_values" in kwargs["kwargs_call"] and "past" in parameter_names:
kwargs["past"] = kwargs["kwargs_call"].pop("past_key_values")

if len(kwargs["kwargs_call"]) > 0:
raise ValueError(
f"The following keyword arguments are not supported by this model: {list(kwargs['kwargs_call'].keys())}."
)

kwargs.pop("kwargs_call")
if has_kwargs:
output["kwargs"] = kwargs.pop("kwargs_call", {})
else:
if len(kwargs["kwargs_call"]) > 0:
raise ValueError(
f"The following keyword arguments are not supported by this model: {list(kwargs['kwargs_call'].keys())}."
)
kwargs.pop("kwargs_call")
Comment on lines +440 to +447
Copy link
Contributor Author

@gante gante Apr 1, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

encoder_decoder models want the kwargs, all other models will discard them (and throw an error if they are not empty)


for k, v in kwargs.items():
if isinstance(v, allowed_types) or v is None:
Expand Down
8 changes: 0 additions & 8 deletions src/transformers/models/albert/modeling_tf_albert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -551,7 +551,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:

if input_ids is not None and inputs_embeds is not None:
Expand Down Expand Up @@ -785,7 +784,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
outputs = self.albert(
input_ids=input_ids,
Expand Down Expand Up @@ -854,7 +852,6 @@ def call(
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
sentence_order_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFAlbertForPreTrainingOutput, Tuple[tf.Tensor]]:
r"""
Return:
Expand Down Expand Up @@ -976,7 +973,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down Expand Up @@ -1064,7 +1060,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
Expand Down Expand Up @@ -1158,7 +1153,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down Expand Up @@ -1244,7 +1238,6 @@ def call(
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
Expand Down Expand Up @@ -1355,7 +1348,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
Expand Down
3 changes: 0 additions & 3 deletions src/transformers/models/bart/modeling_tf_bart.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -679,7 +679,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
"""
Args:
Expand Down Expand Up @@ -834,7 +833,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
Args:
Expand Down Expand Up @@ -1273,7 +1271,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSeq2SeqLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down
9 changes: 0 additions & 9 deletions src/transformers/models/bert/modeling_tf_bert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -737,7 +737,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:

if not self.config.is_decoder:
Expand Down Expand Up @@ -1067,7 +1066,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
encoder_hidden_states (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
Expand Down Expand Up @@ -1174,7 +1172,6 @@ def call(
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
next_sentence_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBertForPreTrainingOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down Expand Up @@ -1302,7 +1299,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down Expand Up @@ -1520,7 +1516,6 @@ def call(
return_dict: Optional[bool] = None,
next_sentence_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFNextSentencePredictorOutput, Tuple[tf.Tensor]]:
r"""
Return:
Expand Down Expand Up @@ -1628,7 +1623,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
Expand Down Expand Up @@ -1723,7 +1717,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
Expand Down Expand Up @@ -1857,7 +1850,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down Expand Up @@ -1949,7 +1941,6 @@ def call(
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
Expand Down
3 changes: 0 additions & 3 deletions src/transformers/models/blenderbot/modeling_tf_blenderbot.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -662,7 +662,6 @@ def call(
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
"""
Args:
Expand Down Expand Up @@ -823,7 +822,6 @@ def call(
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
r"""
Args:
Expand Down Expand Up @@ -1276,7 +1274,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFSeq2SeqLMOutput]:
r"""
labels (`tf.tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down
3 changes: 0 additions & 3 deletions src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -667,7 +667,6 @@ def call(
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
"""
Args:
Expand Down Expand Up @@ -827,7 +826,6 @@ def call(
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
r"""
Args:
Expand Down Expand Up @@ -1253,7 +1251,6 @@ def call(
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFSeq2SeqLMOutput]:
r"""
labels (`tf.tensor` of shape `(batch_size, sequence_length)`, *optional*):
Expand Down
12 changes: 0 additions & 12 deletions src/transformers/models/clip/modeling_tf_clip.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -504,7 +504,6 @@ def call(
output_hidden_states: bool,
return_dict: bool,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
input_shape = shape_list(input_ids)

Expand Down Expand Up @@ -593,7 +592,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
if input_ids is None:
raise ValueError("You have to specify input_ids")
Expand Down Expand Up @@ -632,7 +630,6 @@ def call(
output_hidden_states: bool,
return_dict: bool,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:

embedding_output = self.embeddings(pixel_values=pixel_values)
Expand Down Expand Up @@ -683,7 +680,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:

if pixel_values is None:
Expand Down Expand Up @@ -762,7 +758,6 @@ def get_text_features(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:

if input_ids is None:
Expand Down Expand Up @@ -796,7 +791,6 @@ def get_image_features(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
if pixel_values is None:
raise ValueError("You have to specify pixel_values")
Expand Down Expand Up @@ -826,7 +820,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFCLIPOutput, Tuple[tf.Tensor]]:

if input_ids is None:
Expand Down Expand Up @@ -1058,7 +1051,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
r"""
Returns:
Expand Down Expand Up @@ -1153,7 +1145,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
r"""
Returns:
Expand Down Expand Up @@ -1258,7 +1249,6 @@ def get_text_features(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
r"""
Returns:
Expand Down Expand Up @@ -1297,7 +1287,6 @@ def get_image_features(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
r"""
Returns:
Expand Down Expand Up @@ -1345,7 +1334,6 @@ def call(
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFCLIPOutput, Tuple[tf.Tensor]]:
r"""
Returns:
Expand Down
Loading