how to run with 1 gpu? - Githubissues

megvii-research / MSPN

Multi-Stage Pose Network

334 stars 63 forks source link

Closed chenyanyin closed 5 years ago

chenyanyin commented 5 years ago

thx

megvii-wzc commented 5 years ago

See the detail in the README: "python -m torch.distributed.launch --nproc_per_node=gpu_num train.py". Set the gpu_num=1 means running with 1 gpu.

chenyanyin commented 5 years ago

i got it, thx. Sorry the problem that i met is not because of it, but i solved. thx again.

chenyanyin commented 5 years ago

@megvii-wzc Does the code support continuing training when interrupted? How to do it if support?

fenglinglwb commented 5 years ago

You just need to add another argument -c to specify the path of model that you want to continue.

chenyanyin commented 5 years ago

ok, thx.