cperry-goog
commented
8 months ago You can use Colab in a runtime environment of your choosing with our Docker container or on your own local runtime w/ instructions provided here: https://research.google.com/colaboratory/local-runtimes.html
Keeping this open for the general request of giving users more first class support for custom containers in managed runtimes.
Thank you for providing the Colab service, which allows me to carry out my data analysis work in a convenient and inexpensive way. During my work, I have encountered some inconveniences and I have some nebulous ideas that I hope to discuss with you. My biggest headache is that some commonly used Python packages in my field of research often conflict with the Colab environment, such as the Python version, the Numpy version, etc. I tried using virtualenv to solve this, but found that it seems impossible to activate venv in the Colab notebook. Therefore, I was wondering if it might be possible for Colab to offer more personalized settings, such as the ability to choose the Python version? Or even go a step further and turn Colab into Co-VM, allowing us to run our own Docker images, which seems to be a one-time solution for personalized runtime environment needs? I realize that this may contradict the positioning of the Colab product and overlap more with GCP services. I'm just imagining how wonderful it would be to offer GCP's computing resources to users in the form of Colab (cheap, convenient, no need to consider server architecture, can be quickly activated when needed, and deleted directly after use) for researchers who do not have high-performance servers. The above are some of the problems I encountered in my work and some whimsical ideas I had, and I hope to discuss them with everyone. Regardless, I am grateful for the services provided by Colab.