Deep Learning Support at Indiana University
summary
Indiana University adds a 12-node expansion to the Carbonate cluster with eight nodes sporting dual P100 NVIDIA GPUs and four with dual V100 GPUs, exposes frameworks like TensorFlow PyTorch Caffe Keras CNTK MxNet and provides software support and consulting for researchers, reports measurable performance tuning that improves runtimes on V100 hardware, offers project access policies and 10–75 PB storage tiers, and makes the GPUs and configured stacks available to any qualifying deep learning project.
AI-written summary · about
chapters (17)
- 0:00IU research computing intro
- 2:36Data storage at IU
- 5:07Karst and Carbonate systems
- 7:12Research desktop computing
- 9:02Slate and Data Capacitor
- 10:03Deep learning support challenges
- 11:09GPU specs for deep learning
- 13:21Big Red 200 supercomputer
- 16:15Accessing IU supercomputing
- 18:13IU vs public cloud
- 21:21Why IU uses Cray
- 25:18Deep learning research at IU
- 26:40System utilization policies
- 27:45Pricing and cost comparisons
- 30:54Scientific impact and return on investment
- 31:40Data security policies
- 33:04Supercomputer lifecycle
related talks
25mHAL: Computer System for Scalable Deep Learning
31mHigh-Performance MPI and Deep Learning on OpenPOWER Platform
39mIntel Nervana DevCloud
18mLesson Learned from the Building of a Terascale AI Server Infrastructure
25mHigh-Performance MPI and Deep Learning with Introspection...
49mDistributed Deep Learning
28mA Deep Dive on How To Leverage the NVIDIA GB200 for Ultra-Fast...
44mDeep Learning in Production