fast-autoaugment icon indicating copy to clipboard operation
fast-autoaugment copied to clipboard

Stuck in search.py

Open MaDaJie2706 opened this issue 3 years ago • 5 comments

First of all, thanks you very much for your generous to sharing code public. My problem happen after i changed @ray.remote(num_gpus=4, max_calls=1) and ray.init(redis_address=args.redis) to @ray.remote(num_gpus=1, max_calls=1) and ray.init() in search.py,the running state of the code remains as shown in the image below.The code ran for two hours without making any progress.And it does not return any error messages. Hope that you can answer question soon.Thank you very much! P/s I use "!python3 FastAutoAugment/search.py -c confs/wresnet40x2_cifar.yaml" to run this code. VB2YUAEDVD5R9{HR4@65SRS

MaDaJie2706 avatar Aug 31 '21 04:08 MaDaJie2706

I also have the same problem. Does anyone have a solution?

Linker-Stars avatar Sep 10 '21 01:09 Linker-Stars

me too

licrane avatar Dec 09 '21 13:12 licrane

same problem without using ray cluster framework

BaronWang0130 avatar Apr 11 '22 08:04 BaronWang0130

any solution for this issue?

kolingv avatar Apr 28 '22 14:04 kolingv

I found out that it stucks in infinite loop in search.py 'while True' cycle When i raised KeyboardInterrupt, it shows: изображение

I removed ray cluster's lines so line numbering may differ, but stuck occurs here:

изображение

interestingly that this stuck occurs on different stages - first, when training 2 of 5 model, second, when training first 5 of 5 models has finished.

Deimos-Apollon avatar Jul 26 '23 08:07 Deimos-Apollon