How to setup fastai v0.7 on gcp instance that is setup for 2019's Part 1

Following the excellent setup instructions here, for setting up GCP for Part 1 (2019), running gcloud compute ssh jupyter@my-fastai-instance -- -L 8080:localhost:8080 allows me to start using the correct setup for the course.

However, I also want to do some parts of 2018’s courses. So, following the instructions here Fastai v0.7 install issues thread, I created a conda envrionment for fastai v0.7. Now, after doing source activate fastai, how do I serve the notebooks? For example, if I now do jupyter notebook --port=8888 -no-browser, how do I connect to it from my computer? Is there a gcloud command that’s equivalent to ssh -f -N -L localhost:8888:localhost:8888 xxxxx?


1 Like

Doh. I guess it was kind of obvious. So, in summary, follow the instructions here Fastai v0.7 install issues thread to create a conda environment for fastai v0.7. After it’s created and activated, jupyter notebook --no-browser --port=8888 runs jupyter at port 8888. Then, on one’s own computer, do gcloud compute ssh --zone="us-west2-b" jupyter@my-fastai-instance -- -L 8888:localhost:8888. Go to url localhost:8888/tree to start using Jupyter. Notebooks that use fastai v0.7 should then all work.

1 Like