Note
Some links in this document require Broad Institute access. These are marked with π.
- Macbook: For most daily tasks
- DGX-1 / spirit / oppy servers: For GPU computing FIXME: Now DGX-1 owned by C-lab
- AWS: For specific project needs (use judiciously)
- Broad's compute cluster: For easily parallelizable jobs
- Maintenance details: https://github.com/broadinstitute/ip-chores/issues/25 π
- Quick start guide: https://new.ipwiki.app/dgx-1 π
- We are piloting nixOS http://github.com/leoank/neusis/
- Best for: Easily parallelizable jobs without interactivity or cloud data access needs
- Getting started: https://intranet.broadinstitute.org/node/5811 π
- Support: #bits Slack channel
- Also see #uger and #devnull-gpu-discuss Slack channels for discussions
- Use cases: Projects requiring interactivity, large-scale cloud data access, or specific GPU needs
- Getting started:
- Request an account through CloudLab for $500 in free credits with a 90-day limit, which will allow you to experiment freely
- After initial experimentation, ask Shantanu to set up your project account
- Required reading:
- Beth's talk π and slides π on Imaging Platform AWS Use. This is mostly relevant to Cimini Lab's usage of AWS but the concepts are useful to know.
- AWS Usage Policy π
- We recommend interacting via Skypilot; see this Slack thread π
- Install applications using brew
- Use pixi for project environment management (conda + pip dependencies, GPU support) β see workflows for details
- Use uv for standalone Python tooling outside of project environments
Important
Complete these steps before starting any new project. Most links below require Broad access π.
You will rarely need to initialize a project yourself, but here are the instructions if needed:
- Create Imaging Platform Project List Item π
- Send startup email π - See this blog post for context
- Continue with "All Projects" steps below
- Find
project_tagandproject_name_fullin Imaging Platform Project List π - CC img-emailcatcher in email correspondence
- Continue with "All Projects" steps below
- Create GitHub repo following wiki guidelines π for lab notebook
- Create folder in ProjectPlanning π:
YYYY_MM_DD_Project_description - Create S3 folder:
s3://imaging-platform/YYYY_MM_DD_Project_description
- Format:
YYYY_MM_DD_Short_description - Example:
2021_08_23_Batch_effect_correction