VT OpenPOWER Hackathon
summary
Virginia Tech runs a two‑week OpenPOWER hackathon with 52 participants using the PowerAI toolset and Open OnDemand as a one‑click portal to Jupyter notebooks and TensorBoard. Teams port models to GPUs: an actor‑critic RL implementation shows dramatic GPU speedup, a Siamese single‑cell RNA classifier scales across GPUs but suffers unexpected accuracy drops, and physics‑aware GANs generate CFD training data with improving realism while convergence and non‑linear scaling of training time remain challenges.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:26Advanced Research Computing
- 3:29VT OpenPOWER Hackathon
- 7:27High Performance Computing
- 8:46Open OnDemand
- 17:18Team: Like A Gauss
- 18:44Team: CompuOmics
related talks
27mHack for Good (and Profit)
51m"Working Code Wins": Win Big With a Cloud Native Hackathon Start...
22mBridging the Hardware Gap With Code Harnesses on the Hugging Face Kernels Hub
31mThe POWER is with us, let's put it to work!
37mPanel Discussion
46mThe AI-Powered Developer
37mWelcome & Opening Ceremonies
25mDevcontainers for AI/ML Workflow Optimization
