d2l-zh
d2l-zh copied to clipboard
d2l-ai
Add PaddlePaddle Implementation
Initiated evaluation from scratch: http://ci.d2l.ai/blue/organizations/jenkins/d2l-zh/detail/PR-1198/1/pipeline http://ci.d2l.ai/blue/organizations/jenkins/d2l-zh/detail/PR-1198/2/pipeline/ http://ci.d2l.ai/blue/organizations/jenkins/d2l-zh/detail/PR-1198/3/pipeline/
Job d2l-zh/PR-1198/3 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
LGTM
Per http://ci.d2l.ai/blue/organizations/jenkins/d2l-zh/detail/PR-1198/3/pipeline/, it looks that the paddle evaluation runs significantly longer (1h47m) than implementation with other frameworks (pytorch 38m). Could you look into the complete log (click "Show Complete Log") and identify individual notebooks that take longer to evaluate? Can you improve the efficiency?
For example:
- why is there this error:
- In paddle complete log:
[d2lbook:resource.py:L223] INFO Task "Evaluating ./chapter_computer-vision/image-augmentation.md" on CPU [1], GPU [1, 0] is finished in 00:12:19
In pytorch complete log:
[d2lbook:resource.py:L223] INFO Task "Evaluating ./chapter_computer-vision/image-augmentation.md" on CPU [3], GPU [1, 3] is finished in 00:02:36
There can be more examples like the above.
Job d2l-zh/PR-1198/4 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/5 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/6 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Hi, @astonzhang, I would like to know how to make the CI build the whole project again? it seems that the CI only builds the files that I modified.
Job d2l-zh/PR-1198/9 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
we started a CI from scratch. @w5688414
Hi, @cheungdaven Could you please rebuild whole projects, it seems that I can't rerun the whole project by myself.
Job d2l-zh/PR-1198/10 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Hi, @astonzhang , I removed multi gpu training, because paddle doesn't support multi gpu training on notebooks, so it takes about 50mins to build the paddle version's of d2l notebooks.
what's more, I can't reproduce the error on my local machine, it seems that the error won't affect the building procedure, is this a serivous problem? can you give me some suggestions?
--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
No stack trace in paddle, may be caused by external reasons.
----------------------
Error Message Summary:
----------------------
FatalError: `Termination signal` is detected by the operating system.
[TimeInfo: *** Aborted at 1662692867 (unix time) try "date -d @1662692867" if you are using GNU date ***]
[SignalInfo: *** SIGTERM (@0x3e900017699) received by PID 95900 (TID 0x7f2bcf693080) from PID 95897 ***]
Hi, @astonzhang , I removed multi gpu training, because paddle doesn't support multi gpu training on notebooks, so ti takes about 50mins to build the paddle version's of d2l notebooks.
what's more, I can't reproduce the error on my local machine, it seems that the error won't affect the building procedure, is this a serivous problem? can you give me some suggestions?
-------------------------------------- C++ Traceback (most recent call last): -------------------------------------- No stack trace in paddle, may be caused by external reasons. ---------------------- Error Message Summary: ---------------------- FatalError: `Termination signal` is detected by the operating system. [TimeInfo: *** Aborted at 1662692867 (unix time) try "date -d @1662692867" if you are using GNU date ***] [SignalInfo: *** SIGTERM (@0x3e900017699) received by PID 95900 (TID 0x7f2bcf693080) from PID 95897 ***]
Can you follow https://github.com/d2l-ai/d2l-zh/blob/paddle/Jenkinsfile and run
pip install git+https://github.com/d2l-ai/d2l-book
d2lbook build eval --tab paddle
to try reproducing the error?
@xiaotinghe Can you review this PR and make sure it doesn't affect the content within the scope of our forthcoming publication?
Hi, @astonzhang , I reproduced the error, the error has been ignored by setting paddle.disable_signal_handler()， Can you review this pr again?
Hi, @astonzhang , I removed multi gpu training, because paddle doesn't support multi gpu training on notebooks, so ti takes about 50mins to build the paddle version's of d2l notebooks.
-------------------------------------- C++ Traceback (most recent call last): -------------------------------------- No stack trace in paddle, may be caused by external reasons. ---------------------- Error Message Summary: ---------------------- FatalError: `Termination signal` is detected by the operating system. [TimeInfo: *** Aborted at 1662692867 (unix time) try "date -d @1662692867" if you are using GNU date ***] [SignalInfo: *** SIGTERM (@0x3e900017699) received by PID 95900 (TID 0x7f2bcf693080) from PID 95897 ***]Can you follow https://github.com/d2l-ai/d2l-zh/blob/paddle/Jenkinsfile and run
pip install git+https://github.com/d2l-ai/d2l-book d2lbook build eval --tab paddleto try reproducing the error?
Job d2l-zh/PR-1198/12 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/11 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
@astonzhang @cheungdaven @xiaotinghe Hi，thank you very much for reviewing our submission, could we know the progress of review work? Looking forward to your reply～🤗
Job d2l-zh/PR-1198/13 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/18 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/20 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/21 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/22 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/
Job d2l-zh/PR-1198/23 is complete. Check the results at http://preview.d2l.ai/d2l-zh/PR-1198/