Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Xd/dev #6

Open
wants to merge 153 commits into
base: master
Choose a base branch
from
Open

Xd/dev #6

wants to merge 153 commits into from

Conversation

xiaoda99
Copy link
Collaborator

RT

thomwolf and others added 30 commits November 30, 2018 13:47
fix typo in input for masked lm loss function
Should be consistent with the fact that an uncased model is used
Update finetuning example in README adding --do_lower_case
I tried to address huggingface#76
should be correct, but there's likely a more efficient way.
Correct assignement for logits in classifier example
thomwolf and others added 24 commits January 7, 2019 13:01
It was modified similar to `run_classifier.py`, and Fixed to run properly even if without `--do_train` param.
Fix it to run properly even if without `--do_train` param.
…saving

add do_lower_case arg and adjust model saving for lm finetuning.
The "default=True" makes args.do_lower_case always True.

```python
parser.add_argument("--do_lower_case",
                        default=True,
                        action='store_true')
```
dicts are not ordered in Python 3.5 or prior, which is a cause of huggingface#175.
This PR replaces one with a list, to keep its order.
This PR adds missing backslashes in LM Fine-tuning subsection in README.md.
Importing unofficial TF models seems to be working well, at least for me.
This PR resolves huggingface#50.
[bug fix] args.do_lower_case is always True
lm_finetuning compatibility with Python 3.5
run_squad2 Don't save model if do not train
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.