Machine Learning Credits: AWS/Paperspace/Crestle etc

As I get deeper into deep learning and apply my learning’s I am finding that I need to use a lot more compute time and I know that there are various topics on credits but I wanted to start a thread where we can share additional information on what everyone has used in order to gain preferably free :sunglasses: or reduced pricing compute time.

Please share below of what you have recently used that has worked.

I predominantly use AWS and this is what I have used so far that work (please note requirements).

  1. Github Student Pack (

    • $150 AWS credit
    • Requirements: Valid student email
    • Link to existing AWS account, AWS Educate Starter Account is limited to CPU use only
    • AWS F&Q
  2. AWS Activate for startups ( for anyone looking for resources for their venture.
    -$1000 AWS credit
    - AWS Business Essentials Online Training ($600 value)
    - AWS Technical Essentials Online Training ($600 value)
    - Requirements: Working company website
    Matching domain email address
    Will be evaluated by AWS for acceptance


Good idea for a thread.

Google Colab is free for unlimited number of 12 hours sessions (on a K40).

Startups get lots of Azure credits by joining Bizspark.


A thing to note about Google Free k80 instances. Google colab gives only about 500mb of vRAM for the free version. I got excited to use the 24gb vRAM. Did due diligence, was a bit disappointed
( - Question raised by a fellow fastai student.
Can someone please confirm about their experience (if any) with k40 on google colab?

I did Part 1 on Colab. Apart from a few notebooks that were really memory and GPU intensive ( language modelling etc) it worked well for me. I had to tweak around a little bit, but overall it was a good experience at no cost so I definitely recommend it especially for the tweaking around with the data part. The final training on a complete larger dataset can be done on a Paperspace/AWS instance.

@amritv I had a doubt. Did you successfully manage to use the Github student pack/AWS educate credits for GPU use? I attempted to do the same but was limited to CPU instances. I think I made the mistake of creating a new AWS account with AWS educate instead of transferring credits to an existing one.

Cognitive Class (formerly Big Data University) in collaboration with IBM offered 1400$ worth of credits and claimed that this was a no bs offer but, as expected, you can’t use the Bare Metal Server (GPU) compute with the credits. I emailed them regarding the same, will update if I receive a reply.

@keratin, I was successfully able to use the credits for GPU use, however I linked the credits to my existing AWS account. I think the issue is that in your case you probably created a starter account which does not allow for GPU use. Its limited to t2, m4.large and m4.xlarge instances.


Alternatively, one can get $300 credit from Google Cloud. Thanks for @Nok and James Lee


I tried doing that. But GPU was not part of the free $300. I was asked to make a prepayment if I need access to GPU. Were you able to access GPU using the free credits on GCP?

Yes. You will need to upgrade your account, that will require you to put your credit card information. However, you can still use that $300 dollars. I heard a lot of people being asked for prepayment, usually because they ask for a lot of GPUs/multiple region. What I did is ask for 1 GPU and states that it is for education purpose and it went smooth.


By default you don’t have the permission to use a GPU on GCP. You should modify your “quota” in order to get access to GPUs. You can easily request the modification compiling a form and they “approve” your request nearly immediately :slight_smile:

To do so you need to upgrade your account, but the 300$ credit remains until the end of the free trial.

1 Like

I tried doing the same. I requested for 1 GPU and I got a mail asking for prepayment of $35. As per the email, it says I can’t access a GPU without making the prepayment. But just now I checked the project details and I do see a GPU in the resource list. So I have to try that out. Thanks nok for your reply.

I did exactly that. And then I got an email asking for prepayment. I remember checking some days back if I have GPU access and there wasn’t one that’s allocated to my project. But now I see one in the resource list. I guess I got one. Will try running notebook and see if it works. Thanks alessia.
I read your blog on Machine Learning. It’s very well written. Keep writing :+1:

1 Like

I guess I got one.


I read your blog on Machine Learning. It’s very well written. Keep writing :+1:

Thank you for your feedback!

1 Like