Exception: process 0 terminated with signal SIGKILL`

Hi,

dependency:
pytorch 1.4.0,
CUDA 10.2
Pytorch_encoding master branch

the following code is run on single GPU( GeForce RTX 2080, 8GB)：
CUDA_VISIBLE_DEVICES=0 python train_dist.py --dataset PContext --model EncNet --aux --se-loss

the following issue happened:

Using poly LR scheduler with warm-up epochs of 0! Starting Epoch: 0 Total Epoches: 80 Traceback (most recent call last): File "train_dist.py", line 319, in <module> main() File "train_dist.py", line 148, in main mp.spawn(main_worker, nprocs=ngpus_per_node, args=(ngpus_per_node, args)) File "/home/qyuan/anaconda3/envs/pytorch_encoding_interpreter/lib/python3.6/site-packages/torch/multiprocessing/spawn.py", line 171, in spawn while not spawn_context.join(): File "/home/qyuan/anaconda3/envs/pytorch_encoding_interpreter/lib/python3.6/site-packages/torch/multiprocessing/spawn.py", line 107, in join (error_index, name) Exception: process 0 terminated with signal SIGKILL

I have read all the issues related to it, but the issue is still there,could you please give me some suggestions? thanks for your time!

Reproduce the command in the issue with PyTorch 1.4.0, CUDA 10.2, and a single GeForce RTX 2080. Start in train_dist.py around main() lines 148 and 319, follow the mp.spawn path into main_worker, and use the traceback and process termination to identify the cause; done means training completes without process 0 receiving SIGKILL.