You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I was trying to run the predictions according to the steps, unfortunately, there is some version mismatch problem which I counld not solve, I install all the packages according to the requirement.txt and set the python version to 3.7.11. But when I run the ProtBERT_feature_generator.py, following problem occured.
Traceback (most recent call last):
File "ProtBERT_feature_generator.py", line 4, in <module>
from transformers import BertModel, BertTokenizer
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/__init__.y", line 43, in <module>
from . import dependency_versions_check
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/dependenc_versions_check.py", line 41, in <module>
require_version_core(deps[pkg])
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/utils/verions.py", line 94, in require_version_core
return require_version(requirement, hint)
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/utils/verions.py", line 85, in require_version
if want_ver is not None and not ops[op](version.parse(got_ver), version.pars(want_ver)):
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/packaging/version.py",line 54, in parse
return Version(version)
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/packaging/version.py",line 200, in __init__
raise InvalidVersion(f"Invalid version: '{version}'")
packaging.version.InvalidVersion: Invalid version: '0.10.1,<0.11'
seems to be some version mismatch problem. I tried to uninstall the tokenizers package, so that the program can go on. and the following problem appeared.
Traceback (most recent call last):
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/modeling_utils.py", line 1038, in from_pretrained
state_dict = torch.load(resolved_archive_file, map_location="cpu")
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/torch/serialization.py", line 386, in load
return _load(f, map_location, pickle_module, **pickle_load_args)
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/torch/serialization.py", line 580, in _load
deserialized_objects[key]._set_from_file(f, offset, f_should_read_directly)
RuntimeError: unexpected EOF, expected 3715186 more bytes. The file might be corrupted.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "ProtBERT_feature_generator.py", line 67, in <module>
generate_protbert_features(file)
File "ProtBERT_feature_generator.py", line 24, in generate_protbert_features
model = BertModel.from_pretrained(path)
File "/opt/conda/envs/ensem/lib/python3.7/site-packages/transformers/modeling_utils.py", line 1041, in from_pretrained
f"Unable to load weights from pytorch checkpoint file for '{pretrained_model_name_or_path}' "
OSError: Unable to load weights from pytorch checkpoint file for './' at './pytorch_model.bin'If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.
I have no idea why this problem happens, is the pretrained model a tf model? It would be nice if some help can be offered. Thanks in advance.
The text was updated successfully, but these errors were encountered:
we tested the scripts and no such traceback reported; please check or rebuild your virtual environment to make sure the "transformers" package (version = 4.3.3) is correctly installed.
Hello,
I was trying to run the predictions according to the steps, unfortunately, there is some version mismatch problem which I counld not solve, I install all the packages according to the requirement.txt and set the python version to 3.7.11. But when I run the ProtBERT_feature_generator.py, following problem occured.
seems to be some version mismatch problem. I tried to uninstall the tokenizers package, so that the program can go on. and the following problem appeared.
I have no idea why this problem happens, is the pretrained model a tf model? It would be nice if some help can be offered. Thanks in advance.
The text was updated successfully, but these errors were encountered: