CODEWITHSUNDEEP
google-cloud-platformpytorchgoogle-cloud-ml
anmolagrawal
anmolagrawal

Reputation: 203

Can we deploy Pytorch models to Google Cloud ML Engine? If so, how?

I searched if anybody has. Somebody has provided a work around but couldn't confirm its working. Here are the links -

How to setup pytorch in google-cloud-ml

https://discuss.pytorch.org/t/pytorch-in-google-cloud-ml-engine/12818

Currently, I am using Floydhub, but it is good to start with. I work mostly with Google Cloud products, so want to keep my deep learning projects as close to it as possible.

Upvotes: 0

Views: 1359

Answers (1)

rhaertel80
rhaertel80

Reputation: 8399

The answer you referred to will likely work (I've also not fully tested it). The Cloud ML Engine team is working hard to simplify this.

In the meantime, if you prefer working with a VM (just remember to shut it off when not in use), follow instructions on this page. For example,

export IMAGE_FAMILY="pytorch-latest-cu91"
export ZONE="us-west1-b"
export INSTANCE_NAME="my-instance"

gcloud compute instances create $INSTANCE_NAME \
  --zone=$ZONE \
  --image-family=$IMAGE_FAMILY \
  --image-project=deeplearning-platform-release \
  --maintenance-policy=TERMINATE \
  --accelerator='type=nvidia-tesla-v100,count=1' \
  --metadata='install-nvidia-driver=True'

Upvotes: 1

Related Questions