SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

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

Browse Source 
Revert "Fix gradient checkpointing + fp16 autocast for most models (#24247)"

This reverts commit 285a48011d.
This commit is contained in:
Younes Belkada
2023-06-22 16:11:27 +02:00
committed by GitHub
parent ebb62e8880
commit 3ce3385c47
179 changed files with 271 additions and 836 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/models/vit_mae/test_modeling_vit_mae.py
View File

@@ -208,12 +208,6 @@ class ViTMAEModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_for_pretraining(*config_and_inputs)
@unittest.skip(
reason="The model does not support GC + autocast + fp16: https://github.com/huggingface/transformers/pull/24247"
)
def test_training_gradient_checkpointing_autocast(self):
pass
# overwrite from common since ViTMAEForPretraining has random masking, we need to fix the noise
# to generate masks during test
def check_pt_tf_models(self, tf_model, pt_model, pt_inputs_dict):