riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-08-03 03:31:05 +06:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Allow passing encoder_ouputs as tuple to EncoderDecoder Models (#16814)

Browse Source 
* Add passing encoder_outputs as tuple to existing test

* Add check for tuple

* Add check for tuple also for speech and vision

Co-authored-by: jsnfly <jsnfly@gmx.de>
This commit is contained in:
jsnfly 2022-04-18 19:49:58 +02:00 committed by GitHub
parent 51fa7191b1
commit 51e0ebedcb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 25 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/transformers/models/encoder_decoder/modeling_encoder_decoder.py
View File

@ -22,7 +22,7 @@ from torch import nn
from torch.nn import CrossEntropyLoss
from ...configuration_utils import PretrainedConfig
from ...modeling_outputs import Seq2SeqLMOutput
from ...modeling_outputs import BaseModelOutput, Seq2SeqLMOutput
from ...modeling_utils import PreTrainedModel
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from ..auto.configuration_auto import AutoConfig
@ -494,6 +494,8 @@ class EncoderDecoderModel(PreTrainedModel):
return_dict=return_dict,
**kwargs_encoder,
)
elif isinstance(encoder_outputs, tuple):
encoder_outputs = BaseModelOutput(*encoder_outputs)
encoder_hidden_states = encoder_outputs[0]

4
src/transformers/models/speech_encoder_decoder/modeling_speech_encoder_decoder.py
View File

@ -22,7 +22,7 @@ from torch import nn
from torch.nn import CrossEntropyLoss
from ...configuration_utils import PretrainedConfig
from ...modeling_outputs import Seq2SeqLMOutput
from ...modeling_outputs import BaseModelOutput, Seq2SeqLMOutput
from ...modeling_utils import PreTrainedModel
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from ..auto.configuration_auto import AutoConfig
@ -514,6 +514,8 @@ class SpeechEncoderDecoderModel(PreTrainedModel):
return_dict=return_dict,
**kwargs_encoder,
)
elif isinstance(encoder_outputs, tuple):
encoder_outputs = BaseModelOutput(*encoder_outputs)
encoder_hidden_states = encoder_outputs[0]

4
src/transformers/models/vision_encoder_decoder/modeling_vision_encoder_decoder.py
View File

@ -22,7 +22,7 @@ from torch import nn
from torch.nn import CrossEntropyLoss
from ...configuration_utils import PretrainedConfig
from ...modeling_outputs import Seq2SeqLMOutput
from ...modeling_outputs import BaseModelOutput, Seq2SeqLMOutput
from ...modeling_utils import PreTrainedModel
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from ..auto.configuration_auto import AutoConfig
@ -466,6 +466,8 @@ class VisionEncoderDecoderModel(PreTrainedModel):
return_dict=return_dict,
**kwargs_encoder,
)
elif isinstance(encoder_outputs, tuple):
encoder_outputs = BaseModelOutput(*encoder_outputs)
encoder_hidden_states = encoder_outputs[0]

16
tests/encoder_decoder/test_modeling_encoder_decoder.py
View File

@ -142,6 +142,22 @@ class EncoderDecoderMixin:
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
)
# Test passing encoder_outputs as tuple.
encoder_outputs = (encoder_hidden_states,)
outputs_encoder_decoder = enc_dec_model(
encoder_outputs=encoder_outputs,
decoder_input_ids=decoder_input_ids,
attention_mask=attention_mask,
decoder_attention_mask=decoder_attention_mask,
)
self.assertEqual(
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
)
self.assertEqual(
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
)
def check_encoder_decoder_model_from_pretrained_using_model_paths(
self,
config,