Pipeline error when creating a model without a model card json file (on Windows)

AlecS12 commented 4 years ago

🐛 Bug

Model I am using (Bert, XLNet....): Bert Language I am using the model on (English, Chinese....): English The problem arise when using:

[ ] the official example scripts: (give details)

[x] my own modified scripts: (give details)


import sys
from transformers import pipeline

if name == 'main': print("start") nlp_ft = pipeline('question-answering', model=r'C:\Users\a652726\PycharmProjects\src\data\raw\qa\wwm-bert-uncased-finetuned-squad', tokenizer='bert-large-uncased')


Results in the ValueError: no modelcard.json file (which i do not have)
My  fix (hack):
In modelcard.py replace (line 164):

```python
        except EnvironmentError:
            if pretrained_model_name_or_path in ALL_PRETRAINED_CONFIG_ARCHIVE_MAP:

with

        except (EnvironmentError, ValueError):
            if pretrained_model_name_or_path in ALL_PRETRAINED_CONFIG_ARCHIVE_MAP:

This results in:

            logger.warning("Creating an empty model card.")

And everything works fine after this.

The tasks I am working on is:

[ ] an official GLUE/SQUaD task: (give the name)
[x ] my own task or dataset: (give details) Changing QA pipeline to work on a fixed set of spans (like multiple choice QA task, or classification)

To Reproduce

Steps to reproduce the behavior:

1. 2. 3.

Expected behavior

Environment

OS: Windows
Python version: 3.7
PyTorch version: 1.3
PyTorch Transformers version (or branch): master (post 2.3)
Using GPU ? no
Distributed or parallel setup ? no
Any other relevant information:

Additional context

BramVanroy commented 4 years ago

Please format your post correctly by using code blocks. https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

huggingface / transformers