I am just starting getting into deep learning with tf.keras. I am at the point where I have to decide where I want to develop. The thesis project will be timeseries prediction. My options are PyCharm on Macbook Air M1 2020 or a 2013 4th Gen Intel i5 Linux desktop and Google Colab (please let me know if there are others).

So the question I have now is which one is faster/better suited for my puropses. M1 got hyped a lot so I thought the M1 would savage my desktop (and acutally the hype biased my purchase decision), but well its only slightly better (like 1.2-1.5x faster in my cifar10 benchmark) and I wonder if its worth the effective 1-2 GB of RAM left on MacOS vs the ~14 GB on my Linux machine. Further there is Colab and I can't really tell which one will win the race, since Colab limits resources by demand but also allows distributed fit on cloud TPUs, which would introduce some extra coding efforts. Then again I have to say: so does ML on Apple Silicon, which comes with a handful of limitations, a peculiar MiniConda setup, a lot of issues (also severe ones, like training errors etc., problems which I would not even recognize) which are actually not really being worked on.

Is from the perspective of professional data scientists, which I hope to find here, a clear indication on which I should choose?

  • 1
    $\begingroup$ I would opt for using Google Colab, since this will give you a GPU to use. While I am not fully up to date with the M1 performance but have heard good things, using a GPU should allow you to train models much faster than using just a CPU. $\endgroup$
    – Oxbowerce
    Apr 30 at 9:40
  • $\begingroup$ You should use colab or similar environments. If you are willing to pay you can try floydhub.com as well. $\endgroup$
    – iva123
    Apr 30 at 15:02

Better to use Collab indeed. Kaggle also provides notebooks with 38h GPU and also 30 hours of TPU per week you might want to have a look at that as well (plus Kaggle allows you to use your GCP credentials so you can link private google cloud storage buckets to your Kaggle notebook). On Kaggle you will also find plenty of public notebooks that can be of great help


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.