I try to use the google machine learning engine to train my tensorflow model. I want to use the free tier.
When I set the configuration and create new compute engine instance ( Google compute engine), I try to add the GPU too. But this gives me an error because the Google cloud give just $300 credit free tier without GPU.
Is there any way to use the GPU in free tier?
As per the Google's doc
New projects and Free Trial accounts do not receive GPU quota by default.
You must have GPU quota before you can create instances with GPUs.
All free tiers users are initiated with 0 GPU, which they will have to submit a request to increase the quota. However, to file a request, one has to upgrade his account to a paid account.
10 minutes ago, I am still in the free tier and all my projects will be billed for the $300 free trial promotion credits.
I had to upgrade my account to submit a request for GPUs. A minute later my request was approved. When I try to create an instance again, I am charged according to the price sheet.
Updates
I have just reached out to a Google Cloud Support member and here's the result.
After an account has been upgraded, the unconsumed promotion $300 credits will be consumed first. Once the credits have been consumed, the charges will automatically then charged to your payment method on file. Since you have upgraded your billing account, you will not be notified that your free trial credits are consumed and that charges will be taken on your credit card. However, you can actually monitor your charges on your account by simply logging in to your console and check usage in the billing section.
In short, you may upgrade your account first and train your machine learning model and monitor the charges in the meantime to avoid unintended charges.
Meanwhile, Google has just released out the Colab sometime earlier to the general public with a noble goal of dissemination of machine learning education and research.
You can use GPU as a backend for free for 12 hours at a time.
The GPU used in the backend is K80(at this moment).
The 12-hour limit is for a continuous assignment of VM. It means we can use GPU compute even after the end of 12 hours by connecting to a different VM.
This essentially means you can train your ML model for free! Hurray!
Hope my findings help.
Update as on 3/14/18:
Google Cloud ML Engine now allows you to use GPUs with your jobs in the free tier. I am currently training using the BASIC_GPU scale tier, which uses a single standard-gpu instance as master machine. I haven't yet tried using cluster of multiple GPUs though, will update when I find information on this.
Edit: If you are learning and want to experiment with your Machine/Deep learning models, you can give Google's Colab a try, which provides free GPU access for the same purpose.
You can't choose a GPU because, as a free trial your GPU quota is set to 0. Simply go to the quotas panel on your Google Cloud Console, then you will be able to edit you quota there to some value other than 0 by clicking on "Edit quotas" and selecting the kind of GPU you want to use.
