Skip to content
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.

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

Revert "Fix gradient checkpointing + fp16 autocast for most models" #24420

Merged
merged 1 commit into from
Jun 22, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
6 changes: 1 addition & 5 deletions src/transformers/models/albert/modeling_albert.py
Original file line number Diff line number Diff line change
Expand Up @@ -34,11 +34,7 @@
TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import (
ModelOutput,
add_code_sample_docstrings,
Expand Down
9 changes: 2 additions & 7 deletions src/transformers/models/align/modeling_align.py
Original file line number Diff line number Diff line change
Expand Up @@ -30,12 +30,7 @@
BaseModelOutputWithPoolingAndNoAttention,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
torch_custom_checkpointing,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import (
ModelOutput,
add_start_docstrings,
Expand Down Expand Up @@ -1105,7 +1100,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
11 changes: 3 additions & 8 deletions src/transformers/models/altclip/modeling_altclip.py
Original file line number Diff line number Diff line change
Expand Up @@ -30,12 +30,7 @@
BaseModelOutputWithPoolingAndProjection,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
torch_custom_checkpointing,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import ModelOutput, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from .configuration_altclip import AltCLIPConfig, AltCLIPTextConfig, AltCLIPVisionConfig

Expand Down Expand Up @@ -656,7 +651,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -970,7 +965,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@
from ...activations import ACT2FN
from ...modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling, SequenceClassifierOutput
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import find_pruneable_heads_and_indices, prune_linear_layer, torch_custom_checkpointing
from ...pytorch_utils import find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, logging
from .configuration_audio_spectrogram_transformer import ASTConfig

Expand Down Expand Up @@ -343,7 +343,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
layer_head_mask,
Expand Down
5 changes: 2 additions & 3 deletions src/transformers/models/autoformer/modeling_autoformer.py
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@
Seq2SeqTSPredictionOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...time_series_utils import NegativeBinomialOutput, NormalOutput, StudentTOutput
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from .configuration_autoformer import AutoformerConfig
Expand Down Expand Up @@ -1211,7 +1210,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -1429,7 +1428,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
5 changes: 2 additions & 3 deletions src/transformers/models/bart/modeling_bart.py
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,6 @@
Seq2SeqSequenceClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
add_code_sample_docstrings,
add_end_docstrings,
Expand Down Expand Up @@ -850,7 +849,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -1106,7 +1105,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
4 changes: 2 additions & 2 deletions src/transformers/models/beit/modeling_beit.py
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@
SemanticSegmenterOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import find_pruneable_heads_and_indices, meshgrid, prune_linear_layer, torch_custom_checkpointing
from ...pytorch_utils import find_pruneable_heads_and_indices, meshgrid, prune_linear_layer
from ...utils import (
add_code_sample_docstrings,
add_start_docstrings,
Expand Down Expand Up @@ -517,7 +517,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
layer_head_mask,
Expand Down
9 changes: 2 additions & 7 deletions src/transformers/models/bert/modeling_bert.py
Original file line number Diff line number Diff line change
Expand Up @@ -40,12 +40,7 @@
TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
torch_custom_checkpointing,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import (
ModelOutput,
add_code_sample_docstrings,
Expand Down Expand Up @@ -603,7 +598,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -25,12 +25,7 @@
from ...activations import ACT2FN
from ...modeling_outputs import BaseModelOutputWithPastAndCrossAttentions, CausalLMOutputWithCrossAttentions
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
torch_custom_checkpointing,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import (
add_code_sample_docstrings,
add_start_docstrings,
Expand Down Expand Up @@ -413,7 +408,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
4 changes: 2 additions & 2 deletions src/transformers/models/big_bird/modeling_big_bird.py
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@
TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import apply_chunking_to_forward, torch_custom_checkpointing
from ...pytorch_utils import apply_chunking_to_forward
from ...utils import (
ModelOutput,
add_code_sample_docstrings,
Expand Down Expand Up @@ -1622,7 +1622,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,6 @@
Seq2SeqSequenceClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
add_code_sample_docstrings,
add_end_docstrings,
Expand Down Expand Up @@ -1946,7 +1945,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -2292,7 +2291,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
3 changes: 1 addition & 2 deletions src/transformers/models/biogpt/modeling_biogpt.py
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,6 @@
TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
add_code_sample_docstrings,
add_start_docstrings,
Expand Down Expand Up @@ -595,7 +594,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
5 changes: 2 additions & 3 deletions src/transformers/models/blenderbot/modeling_blenderbot.py
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,6 @@
Seq2SeqModelOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
add_end_docstrings,
add_start_docstrings,
Expand Down Expand Up @@ -780,7 +779,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -1035,7 +1034,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@
Seq2SeqModelOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
add_end_docstrings,
add_start_docstrings,
Expand Down Expand Up @@ -778,7 +777,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -1032,7 +1031,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(decoder_layer),
hidden_states,
attention_mask,
Expand Down
3 changes: 1 addition & 2 deletions src/transformers/models/blip/modeling_blip.py
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,6 @@
from ...activations import ACT2FN
from ...modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import (
ModelOutput,
add_start_docstrings,
Expand Down Expand Up @@ -621,7 +620,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down
3 changes: 1 addition & 2 deletions src/transformers/models/blip/modeling_blip_text.py
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@
find_pruneable_heads_and_indices,
prune_linear_layer,
)
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import logging
from .configuration_blip import BlipTextConfig

Expand Down Expand Up @@ -428,7 +427,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
11 changes: 3 additions & 8 deletions src/transformers/models/blip_2/modeling_blip_2.py
Original file line number Diff line number Diff line change
Expand Up @@ -31,12 +31,7 @@
BaseModelOutputWithPoolingAndCrossAttentions,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import (
apply_chunking_to_forward,
find_pruneable_heads_and_indices,
prune_linear_layer,
torch_custom_checkpointing,
)
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import (
ModelOutput,
add_start_docstrings,
Expand Down Expand Up @@ -497,7 +492,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
Expand Down Expand Up @@ -968,7 +963,7 @@ def custom_forward(*inputs):

return custom_forward

layer_outputs = torch_custom_checkpointing(
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
Expand Down
3 changes: 1 addition & 2 deletions src/transformers/models/bloom/modeling_bloom.py
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,6 @@
TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import torch_custom_checkpointing
from ...utils import logging
from .configuration_bloom import BloomConfig

Expand Down Expand Up @@ -776,7 +775,7 @@ def custom_forward(*inputs):

return custom_forward

outputs = torch_custom_checkpointing(
outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(block),
hidden_states,
alibi,
Expand Down
Loading