riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-08-02 03:01:07 +06:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Fix generation config for empty state dict (#21630)

Browse Source
This commit is contained in:
Sylvain Gugger 2023-02-14 10:57:28 -05:00 committed by GitHub
parent 317282927d
commit d4ba6e1a0e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 13 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/transformers/modeling_utils.py
View File

@ -2648,7 +2648,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
_from_pipeline=from_pipeline,
**kwargs,
)
except OSError:
except (OSError, TypeError):
logger.info(
"Generation config file not found, using a generation config created from the model config."
)

21
tests/test_modeling_common.py
View File

@ -325,6 +325,18 @@ class ModelTesterMixin:
else:
check_save_load(first, second)
def test_from_pretrained_no_checkpoint(self):
config, _ = self.model_tester.prepare_config_and_inputs_for_common()
for model_class in self.all_model_classes:
model = model_class(config)
state_dict = model.state_dict()
new_model = model_class.from_pretrained(
pretrained_model_name_or_path=None, config=config, state_dict=state_dict
)
for p1, p2 in zip(model.parameters(), new_model.parameters()):
self.assertTrue(torch.equal(p1, p2))
def test_save_load_keys_to_ignore_on_save(self):
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
@ -2776,15 +2788,6 @@ class ModelUtilsTest(TestCasePlus):
BertModel.from_pretrained(TINY_T5)
self.assertTrue("You are using a model of type t5 to instantiate a model of type bert" in cl.out)
def test_model_from_pretrained_no_checkpoint(self):
config = BertConfig.from_pretrained("hf-internal-testing/tiny-random-bert")
model = BertModel(config)
state_dict = model.state_dict()
new_model = BertModel.from_pretrained(pretrained_model_name_or_path=None, config=config, state_dict=state_dict)
for p1, p2 in zip(model.parameters(), new_model.parameters()):
self.assertTrue(torch.equal(p1, p2))
def test_model_from_config_torch_dtype(self):
# test that the model can be instantiated with dtype of user's choice - as long as it's a
# float dtype. To make it happen config.torch_dtype needs to be set before instantiating the