-
Notifications
You must be signed in to change notification settings - Fork 26.4k
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
NER - pl example #3180
NER - pl example #3180
Conversation
shubhamagarwal92
commented
Mar 8, 2020
•
edited
Loading
edited
- Solves most of the issues raised in NER: some issues in PyTorch Lightning example #3159
- Streamlines shell script pipeline
- pl logs related changes
- added in Readme
…ents. 2. unrecognized arguments: save_steps
…3. pass gpu ids compared to number of gpus
This looks great to me. Thanks @shubhamagarwal92. You need to just run the black command, I believe to is |
@srush I ran the BTW, |
That would be because TensorFlow only supports python 3.5-3.7, unfortunately. |
@LysandreJik Thanks. Installs on ubuntu with python 3.6. However, on mac:
Mac specs:
|
Hi @shubhamagarwal92 , thanks for that PR and fixing the issues 👍 I just ran the TEST RESULTS
{'val_loss': tensor(7.0679), 'precision': 0.0, 'recall': 0.0, 'f1': 0}
----------------------------------------------------------------------------------------------------
Testing: 200it [00:07, 27.98it/s] Last lines of the prediction output: der I-OTHderiv
Bibliothek I-OTHderiv
berufen I-OTHderiv
wurde I-OTHderiv
, I-OTHderiv
verließ I-OTHderiv
Gardthausen I-OTHderiv
den I-OTHderiv
Bibliotheksdienst I-OTHderiv
. I-OTHderiv |
Thanks for reporting this. Could you please verify the version of Also could you please check if the results in It works for me as: |
Hi, I'm using the same versions of both Output of $ cat germeval-model/test_results.txt
f1 = 0
precision = 0.0
recall = 0.0
val_loss = tensor(9.4173) But I'm going to test it on another machine :) |
I am also attaching my environment file via Please let me know if this doesn't work. |
@shubhamagarwal92 I think somehow you have the wrong version of our style checks installed. Can you try running under this command?
@LysandreJik we have to fix this, it is really confusing... @stefan-it would love to see your log as well. Could you also try |
Codecov Report
@@ Coverage Diff @@
## master #3180 +/- ##
==========================================
+ Coverage 78.15% 78.16% +0.01%
==========================================
Files 98 98
Lines 16641 16641
==========================================
+ Hits 13006 13008 +2
+ Misses 3635 3633 -2
Continue to review full report at Codecov.
|
@srush I reverted the last 3 style related commits, force-pushed and added a small commit to pass all the checks. Please merge if everything is fine. Also, for
This link is also wrong at contributing.md. Could you please also state the python version in the md. This command
|
Thanks @shubhamagarwal92. Sorry for the annoyance. @LysandreJik this lgtm. |
@srush Happy to help! :) Thanks for approving the PR! |
@shashwath94 I think I've found the reason for the bad evaluation results: I'm using apex and the Do you have any idea, why it is not working using half precision 🤔 |
Ah, I will check with pytorch-lightning. It is plausible we are not integrating correctly with them |
@srush While you are it, could you please check the status of my PR in pl as well. Lightning-AI/pytorch-lightning#1094 Basically, I was observing memory leak on GPU0 if other GPU id (eg. [1]) was provided when running the NER example. AFAIK, the solution is |
* 1. seqeval required by ner pl example. install from examples/requirements. 2. unrecognized arguments: save_steps * pl checkpoint callback filenotfound error: make directory and pass * huggingface#3159 pl checkpoint path difference * 1. Updated Readme for pl 2. pl script now also correct displays logs 3. pass gpu ids compared to number of gpus * Updated results in readme * 1. updated readme 2. removing deprecated pl methods 3. finalizing scripts * comment length check * using deprecated validation_end for stable results * style related changes