So you are a brave Kaldi ASR scientist not afraid of getting your hands wet with some computing infrastructure experience and a lot of computing oomph from cloud computing, but don’t know where to start? This crash course on deploying and running BurrMill is the starting point you’ve been looking for!
This course is hands-on, but what you’ll do is not an exercise: we’ll do everything right from the start, and by the end of it you will have you real, production-quality cluster, ready to experiment with. Our goal is also to give you at least some understanding of the Google Cloud computing infrastructure; some things will be simplified in order to not overwhelm you, but none will be dumbed down. Still, some points have shortcuts, and if you feel that your brain is boiling and the cranium is struggling to contain the steam pressure, feel free to skip them for now, but make sure to return to them later. Everything here lays the foundation for future understanding of the platform, and, by extension, what you are really doing.
The course is split into seven rather long blog posts. Go through them in order, and ask for help if you feel lost: we hang out in the BurrMill Q&A Google Group, and will be glad to see you there! Feel free to join right away!
- A gentle introduction to cloud computing for Kaldiers
- GCP Economic Geography 101
- Rev up your GCP
- Set up BurrMill Infrastructure
- Supercharge SSH
- Create Cluster Definition File
- Cluster Deployment and Power States Control
Keep in mind that BurrMill v0.5-beta is an early beta version, so do not freak out if something doesn’t work right. We’re out there to help, and thank you for being a beta-tester!
Other pointers:
- Project repository: https://github.com/burrmill/burrmill
- All BurrMill blog posts: https://100d.space/cat/bm
- Kaldi Help Google Group: https://groups.google.com/forum/#!forum/kaldi-help (read how to join)