Code Ocean capsule for sharing analysis of thalamus MERFISH data, primarily in jupyter notebook form.
To create your own working copy on Code Ocean, select Duplicate from the Capsule menu and use the default option ("Link to git repository"). You can then make your own changes and sync back and forth from our shared github repository (https://github.com/AllenNeuralDynamics/thalamus-merfish-analysis) via git. You can use git functions either in the Reproducibility panel on the right-hand side of the capsule view, or within the cloud workstation (more flexibility).
As a package: install the core functionality as a package directly from github via pip: pip install git+https://github.com/AllenNeuralDynamics/thalamus-merfish-analysis
As a streamlit app: start a Code Ocean cloud workstation in streamlit mode (or from the terminal on another cloud workstation, run streamlit run /code/streamlit_app.py)
Locally: you can clone directly from github to your personal machine, but data assets will not be available. Code Ocean also has an option to download the capsule with data assets, but this will be very large if you're not careful!
We are documenting the data assets used by the capsule by including their name and ID in environment/data_assets.yml. If a data asset is not broadly used, but required for a specific notebook or use case, include the entry commented out with an explanatory comment above. The listed data assets can then be attached from a cloud workstation using the API, by running python environment/attach_data_assets.py (ideally this could run in postInstall, but that's not currently possible).
To use this API call though, you'll need to generate and attach a CO API key (one time only):
We'll generally plan to use a single branch, but feel free to create branches as needed for your own work, and merge back to the master branch to share. Do not attempt to "force push" in git (git push -f), as this seems to be incompatible with how Code Ocean interfaces with git.
Feel free to make minor environment changes, but make sure they work by testing in a cloud workstation before you commit/push. If you're making a dramatic change (installing something with a ton of dependencies) that may be an indication you should create a separate capsule for that work.
Your capsule starts with the folders /code, /data, and /results. See our help article on Paths for more information.
You can upload files to the /code or /data folders using the buttons at the top of the left-side Files pane.
Any plots, figures, and results data should be saved in the /results directory. At the end of each run, these files will appear in the right-side Reproducibility pane, where you can view and download them. When you publish a capsule, the most recent set of results will be preserved as part of the capsule.
Click Environment on the left to find a computational environment to accommodate your software (languages, frameworks) or hardware (GPU) requirements. You can then further customize the environment by installing additional packages. The changes you make will be applied the next time your capsule runs. See our help articles on the computational environment for more information.
The next time you run your capsule after making changes to any part of the environment, a custom environment will be built and cached for future runs.
When you publish a capsule, its computational environment will be preserved with it, thereby ensuring computational reproducibility.
If you run into any issues or have any questions, we're here to help. You can reach out to us on on live chat, or, if you prefer, write to support@codeocean.com.
Use the Help menu above to explore the different sections of our knowledge base, with answers to many common questions.