d2l-ai / d2l-en

Interactive deep learning book with multi-framework code, math, and discussions. Adopted at 500 universities from 70 countries including Stanford, MIT, Harvard, and Cambridge.
https://D2L.ai
Other
23.92k stars 4.36k forks source link

PR #1282 is failing to pass the ci. #1286

Closed abhinavsp0730 closed 4 years ago

abhinavsp0730 commented 4 years ago

@astonzhang @AnirudhDagar @terrytangyuan

Issue.

The PR #1282 is failing to pass the ci. You, can find the error here.

Code to reproduce the error.

https://colab.research.google.com/drive/1zaW6HrrhZD1ntx_W4vxpb8Q0KSbc7mk4?usp=sharing

Details.

I'm not able to use train_ch6 more than one time in a notebook. While executing the train_ch6 function 2nd time I'm getting error. For executing train_ch6 2nd time I've to execute the cell which is having the network architecture once again.

PS: Please refer to the colab notebook for reproducing the error.

terrytangyuan commented 4 years ago

From a quick glance, the variable net might get mis-used/updated inside train(). Could you look into it?

abhinavsp0730 commented 4 years ago

I'm not able to find anyway to run the Train_ch6 more than one.

The only Idea I'm having is to make new variable every time.

But, It'll look ugly. Would you please suggest me a better solution.

On Tue 4 Aug, 2020, 12:10 AM Yuan Tang, notifications@github.com wrote:

The variable net might get referenced/updated inside train(). Could you look into it?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/d2l-ai/d2l-en/issues/1286#issuecomment-668180031, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKM6BK54F4EREOET2GDEKO3R64ACRANCNFSM4PTQB2IQ .

abhinavsp0730 commented 4 years ago

I guess I've to do something like this inside train()

net = trained_net
 del net
return trained_net
abhinavsp0730 commented 4 years ago

update: The issue has been fixed. Thanks, @terrytangyuan for giving me the cause for the error and providing me the correct path to solve it.

closing the issue.