Create a new cloud project here.
- You need a google account to log in.
- When entering the above link, click
+CREATE PROJECT, you will go to a new page. - Accept the term of service.
- Enter the project name, such as 'csc249'.
- Click
CREATE. - There is a $300 promotion at the top of the page, click
ACTIVATE - Filling out the payment page, which requires your payment method. But no worry, you have free $300 promotion and wont't be charged. Finally, click
START MY FREE TRIAL
Since we are using GPUs, we need to acquire the number of GPUs in Quotas. The specific steps are as follows:
-
Upgrade your account as
After updating your account, your billing account will be charged automatically since the $300 promotion runs out. So remember to disable the billing account at here when the remaining amount is insufficient.
-
Filter the metric, edit quota and send a GPU request as follows.
-
Then a conformation email will be sent to your gmail. The request will be approved by email notification within a few hours (2 hours for me).
Having been approved for the GPU request, go to Deep Learning VM and follow the instruction of subsection With one or more GPUs in section Creating a PyTorch Deep Learning VM instance from the Cloud Marketplace. A configuration example is as follow.
-
After clicking Launch on Compute Engine, you can configure the VM as
-
Once the VM has been deployed, the page will update with instructions for accessing the instance as
. Remember the ssh command to the VM. -
Next step is to install cloud sdk described in section Install cloud sdk, so that you can connect to theVM instance.
- Go to the sdk quickstart page and select the suitable OS. Follow the instructions when you can successfully finish
gcloud init. - Run the ssh command copied from step 2 (b). The ssh command template is
gcloud compute ssh --project YOUR_PROJECT_NAME --zone YOUR_ZONE_NAME YOUR_VM_NAME -- -L 8080:localhost:8080 - Now you have done!