saymrwulf/transformers
1
0
Fork 0
mirror of https://github.com/saymrwulf/transformers.git synced 2026-05-15 21:01:19 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix from_pretrained API with config and state_dict (#21542)

Browse source
This commit is contained in:
Sylvain Gugger 2023-02-09 15:44:02 -05:00 committed by GitHub
parent 1efe9c0b24
commit 2020ac4bd6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 13 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
src/transformers/modeling_utils.py
View file

@ -2770,7 +2770,10 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
del state_dict[checkpoint_key]
return mismatched_keys
folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])
if resolved_archive_file is not None:
folder = os.path.sep.join(resolved_archive_file[0].split(os.path.sep)[:-1])
else:
folder = None
if device_map is not None and is_safetensors:
param_device_map = expand_device_map(device_map, original_loaded_keys)

9
tests/test_modeling_common.py
View file

@ -2749,6 +2749,15 @@ 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))
@require_torch
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