conv-social-pooling icon indicating copy to clipboard operation
conv-social-pooling copied to clipboard

Published 20 hours ago • verified publisher icon nachiket92

outputs of train.py

Open zhanghm1819 opened this issue 5 years ago • 3 comments

image

zhanghm1819 avatar Dec 12 '19 05:12 zhanghm1819

When I run train.py given by the author and the second epoch reaches 69.80%, the average training loss suddenly becomes infinity(nan), and then it does not converge again. Does anyone encounter this problem? What should I do?

zhanghm1819 avatar Dec 12 '19 05:12 zhanghm1819

hi, can you tell me how you get the data ? the author seems not to provide. if I can get them, I will do the same example like you. so we may solve the problem.

fgqile avatar Aug 20 '21 03:08 fgqile

hi, can you tell me how you get the data ? the author seems not to provide. if I can get them, I will do the same example like you. so we may solve the problem.

Link to NGSIM Dataset Get US-101-LosAngeles-CA.zip and I-80-Emeryville-CA.zip in attachments

Promisery avatar Sep 03 '21 14:09 Promisery