riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-08-01 02:31:11 +06:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Fix gradient checkpointing bug in Trajectory Transformer (#22125)

Browse Source
This commit is contained in:
Karim Foda 2023-03-13 13:50:02 +00:00 committed by GitHub
parent d0876a095f
commit 4c14c1f47b
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 7 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/transformers/models/trajectory_transformer/modeling_trajectory_transformer.py
View File

@ -533,6 +533,13 @@ class TrajectoryTransformerModel(TrajectoryTransformerPreTrainedModel):
hidden_states = self.drop(token_embeddings + position_embeddings) hidden_states = self.drop(token_embeddings + position_embeddings)
if self.gradient_checkpointing and self.training:
if use_cache:
logger.warning_once(
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
)
use_cache = False
presents = () if use_cache else None presents = () if use_cache else None
all_self_attentions = () if output_attentions else None all_self_attentions = () if output_attentions else None
all_hidden_states = () if output_hidden_states else None all_hidden_states = () if output_hidden_states else None
@ -542,11 +549,6 @@ class TrajectoryTransformerModel(TrajectoryTransformerPreTrainedModel):
all_hidden_states = all_hidden_states + (hidden_states,) all_hidden_states = all_hidden_states + (hidden_states,)
if self.gradient_checkpointing and self.training: if self.gradient_checkpointing and self.training:
if use_cache:
logger.warning_once(
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
)
use_cache = False
def create_custom_forward(module): def create_custom_forward(module):
def custom_forward(*inputs): def custom_forward(*inputs):