I am looking for an online or cloud platform to practice Python and also to practice Machine Learning. Can anyone suggest such resource? Seems like there are a lot of paid services with various tiers of subscription. But, want to seek advise, as I am resuming my journey into programming.
I am currently taking Machine Learning Specialization, and want to pursue this field, AI/Machine Learning/Data Science further.
All suggestions are welcome and appreciated.
Google Colab + Google Drive.
Colab requires no setup effort, has both paid and free version, and offers CPU, GPU and TPU environments with limited runtime. The limitation depends on the version and the environment.
Colab uses a jupyter notebook interface which is very similar to Cousera, only more advanced. You can save your notebook on Google Drive so you will not lose your notebook nor code.
Colab’s storage isn’t persistent, so you lose all data after you end a Colab session or the session is expired due to the runtime limitation. However, it supports mounting a Google Drive so that you can mount it, change your working directory to a path on the Drive, and download data files to the Drive and save any outputs from your code on the Drive.
If you are interested, please search for more about the service and its limitation, also spend some time to experiment and make sure you can create/save/load a notebook, and you can mount a Google drive to persistently store data/outputs.
Given their popularity, there should be a lot of online articles/instructions about them, and with those instructions, it shouldn’t take you more than a day to get started.
For example, you write a notebook that uses data downloaded from the internet and produces some output file to Google Drive. After you shut down that Colab session, if you can reload the same notebook and get back the output files, and if you can view the notebook file and those output files on your Google Drive, then you are good.
Perhaps to try it for a few days and see if it can meet your expectation.
Hi @rmwkwok ,
Thank you so much for your reply. I visited colab.google and spent some time. I will try it using one of the Jupyter notebooks from the course labs and also try to save to Google Drive.
Thank you again.
Course’s notebooks are only designed to be run on the course’s platform. You can’t expect a seamless migration from the course’s platform to Colab, and unless you are eager to research and overcome all those obstacles, I recommend you to stick with the course’s platform for the course’s jupyter notebooks, and use Colab for your own project.
Yes @rmwkwok. I understand that it’s not going to be seamless from labs in the course to Colab. I was looking for my own experimentation/learning purposes, and for my own projects when I get there.
Thanks for your help.
OK! It’s just because I have seen quite a number of examples that learners got stuck here and there to try to run a course’s lab somewhere else. Personally I think it can take too much energy and time than it worths. For example, it might cost us a day or two to try and study a lot of things unrelated to data science just to reproduce what we have seen on Cousera, but by starting from scratch, we could have finished a new model with sklearn in 4 hours by following steps in one online article, mastering new modeling skills.
Although they are some useful skills if we are able to run the lab somewhere else, unless it is absolutely necessary, it will not be the first thing I want to focus on as an ambitous learner in data science.
Good luck to you
Thank you Raymond,
I liked what you said in the last paragraph of your reply. I tend to experiment on my own way trying to recreate or reproduce, a deliberate distraction causing to my ownself. Thanks for highlighting the ‘things to watch’
You are welcom, @Shiva_Kondapalli. I hope things have been going well on your side.
Although we don’t actually know each other, I have a feeling that we could actually be quite similar in terms of having the practice to reproduce. Afterall, my suggestion to follow online articles is not very different in some sense.
If exactly reproducing every single bit of the labs might not be most straightforward, which is sometimes what I observe in this place, selectively and partially reproducing the parts that interest you could be beneficial too, if that does not take too much time.