sample netadapt pruner fails about tinyneuralnetwork

[peterjc123]

@liamsun2019 Fixed. Would you please try again?

[liamsun2019]

I tried the latest version:

python3.6 netadapt_prune.py --batch-size 16 --distributed False

INFO (tinynn.prune.netadapt_pruner) Global Target/Initial FLOPS: 437178624/582904832
INFO (tinynn.prune.netadapt_pruner) Start iteration 1
INFO (tinynn.prune.netadapt_pruner) Init pool process with cuda id 0
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/queues.py", line 234, in _feed
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
File "/usr/local/lib/python3.6/dist-packages/torch/multiprocessing/reductions.py", line 322, in reduce_storage
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 191, in DupFd
File "/usr/lib/python3.6/multiprocessing/resource_sharer.py", line 48, in init
OSError: [Errno 24] Too many open files

Is there any config related to file-open limitations？

[peterjc123]

Please use less workers in your Dataloaders.

[liamsun2019]

I tried that way. But it still fails even setting worker to 1.

[liamsun2019]

I checked the file open number, it's over 670 times for fd dup operations while no close ops are done before the error emerges.

[peterjc123]

@liamsun2019 Could you please use a higher limit? We usually perform the experiment on a server, so we won't encounter this kind of problem.

[liamsun2019]

I understand the environment difference. The prune goes well after I ulimit -n 2048.

[peterjc123]

Ok, glad it worked at your side. I'll close it for now.