Search runs only for 1 iteration

[IlyaTrofimov]

Hi!

I run your code with python search.py --name cifar10 --dataset cifar10 --batch_size 16 --gpu 2

and progress doesn't go further this line for several hours 09/10 11:26:50 AM | Train: [ 1/50] Step 000/781 Loss 2.312 Prec@(1,5) (15.6%, 56.2%)

Can you please check where could be a problem?

[Arui1]

hi,I have met this problem too. Can you solve it?

[IlyaTrofimov]

hi,I have met this problem too. Can you solve it?

I loaded docker with pt.darts and studied versions of all libraries. Then I installed same versions via conda, it helped. I assume some problem with data loader or torchvision version.

[Arui1]

thanks, this problem is due to data loader, it stuck.

[suryakumara]

thanks, this problem is due to data loader, it stuck.

Hi can you explain more about data loader problem.. and how to fix it ?

[Wanli-Jiang]

Try to set num_worker=0 in dataloader.

[shield-coder]

Hi!

I run your code with python search.py --name cifar10 --dataset cifar10 --batch_size 16 --gpu 2

and progress doesn't go further this line for several hours 09/10 11:26:50 AM | Train: [ 1/50] Step 000/781 Loss 2.312 Prec@(1,5) (15.6%, 56.2%)

Can you please check where could be a problem?

I used pip install torchvision==0.2.1 before loading dataset each time （I use 2080，CUDA10.1，pytorch 1.3），it worked. You should run the codes after you use pip (I think it can be seen as a updating) although it is not a fundamental solution

[shield-coder]

pin_memory=False