requirements.txtfile in the root directory of the repo vs. having it in the
bindersubdirectory would make that much difference in build time. I was thinking of adding a comment to that effect but thought I'd best check here first.
Hello, thanks for the great tool! I have a question about a kernel dying issue. I'm trying to host a tutorial on binder, everything runs well except for the last bit where I try to plot this array:
san_pedro_sr_rgb <xarray.DataArray (band: 3, y: 4091, x: 8251)> [101264523 values with dtype=uint16]
I'm able to plot this no problem on an 8gb memory laptop and when I monitor the binder resources with top it looks like there is still a lot of free memory when trying to plot. but the plotting still hangs and eventually the kernel dies on this line
%matplotlib inline san_pedro_sr_rgb.plot.imshow(robust=True)
Any suggestions? Is there an unmet matplotlib dependency I need?
topis not useful because it doesn't understand that inside the container you don't have all the memory of the node available. @jhamman's suggestion is the way to go (or using a subset in the demo and telling people "gotta download it if you want the full monty" ... as a hook for people to be motivated to install stuff locally ;) )
from repo2docker.buildpacks.julia import semver semver.find_semver_match("1.1.0", ["1.0.9", "1.0.11", "1.1.0"]) # -> 1.1.0 semver.find_semver_match("1.1.0", ["1.0.9", "1.0.11", "1.1.0", "1.1.2", "1.2.1"]) # -> 1.2.1
jupyter/scipy-notebookstack and more.
git clonein a
startconfiguration file to pull in the latest version and launch time.
requirements.txt, see here and here. If it gets more complex and you need some that pip cannot install, you can move to using an
environment.ymlfile that allows mixing conda and pip installable dependencies. See here and examples here and here. If that isn't sufficient, you can elevate things even further. For example, often I'll also need to additional use
apt.txtif it is something
apt-getusually. And sometimes you'll also need add in use of
start. All those are listed here. So while you can use a Dockerfile, generally you don't need one and will have an easier time with the standard repo2docker configuration files.