3

I am currently running/training MAchine learning models that are very GPU expensive, Google Colab Pro is not giving me enough GPU/RAM

Is there any alternatives with better GPU and more RAM than Google Colab Pro??

The Dan
  • 173
  • 1
  • 7
  • 2
    IMO this question is not appropriate to this forum, as it borders on opinion and non-unique answers – Nikos M. May 05 '21 at 16:48
  • 5
    You can rent compute in any cloud provider with whatever hardware requirements you may have, and then launch a jupyter server there. – noe May 05 '21 at 17:14
  • Which forum should I use here in stackexchange? Let me know and I delete my post @NikosM. – The Dan May 05 '21 at 17:22
  • 1
    I think noe's comment already provided the best answer. All stackexchnage sites I am aware of, reject questions which do not have specific unique answers and border on opinion – Nikos M. May 05 '21 at 17:23
  • I agree with Nikos, imho the question is not objectively answerable. Are you looking for commerical cloud providers with specific features? If yes which features precisely? As it stands the question could even be answered with "buy whatever hardware you need and build your own high performance system", but probably that wouldn't be satisfying for anybody. Fyi AWS probably provides anything you might need (for a price of course). – Erwan May 05 '21 at 22:09
  • Well basically what I need is GPU power like a NVIDIA GeForce RTX 3090 and at least 64GB RAM – The Dan May 06 '21 at 01:24

3 Answers3

3

Here are a few options:

  • Commercial cloud provider: AWS is the go-to cloud provider for virtually any need, they have options for everything. Check out their options for Deep Learning.
  • You could also buy your own hardware
  • Finally from a data science perspective there is also the option of modifying the parameters of the model and/or data so that the training fits with the hardware limitations.
Erwan
  • 24,823
  • 3
  • 13
  • 34
2

Try Kaggle, they offer 30 hours of TPU per week which might help if you are working with tensorflow or torch (TPUs have 128GB of memory). What you should try when running into memory issues is to use half precision floats which reduces memory requirements.

RonsenbergVI
  • 979
  • 3
  • 10
0

You should try paperspace gradient notebooks. You can get access to them for as low as 8$/month (pro version) and also get access to unlimited A4000 GPU which is really powerful. Also, unlike colab it doesn't disconnect after 12 hours. Also, you get larger RAM/storage and various other perks. I used it for a large-scale deep learning project and it gave me far fewer headaches as compared to kaggle or colab.

https://gradient.run/

Check it out!

Aryan Philip
  • 11