"There's only one way in this world to achieve true happiness, and that is to express yourself with all your skill and enthusiasm in a career that appeals to you more than any other. In such a career, you feel a sense of purpose and achievement...It's not work." (published 1949)
Help us spread the word about the CMMRS 2021 (https://t.co/A6KxHEcGPA), a research summer school for undergrads and master students organized by MPI, Cornell and UMD! The call for applications closes March 15th!
The 2nd episode of "How Are Award-winning Systems Research Artifacts Prepared?" is out. @arpanbg shows that we have no excuse not to prepare artifacts for distributed and/or heterogeneous systems s:-|
Congratulations OSDI ’20 Distinguished Artifact Award Winner "Serving DNNs like Clockwork: Performance Predictability from the Bottom Up" (A. Gujarati, R. Karimi, S. Alzayat, W. Hao, A. Kaufmann, Y. Vigfusson, J. Mace)! #osdi20
Today's systems seminar talk was by @arpanbg and @mpi_jcmace, on Clockwork, their new system for predictable DNN inference over shared GPUs: https://t.co/zSXSHcyDAA
Key take: meeting inference SLOs is easy if the serving infrastructure around the actual model is predictable
#jobs Looking for two intermediate Scala FP developers (North American time zones, remote OK) and one junior C/C++ developer (no tz restriction, remote OK). Effective immediately.
Please pass along.
Delighted to discover this morning that our OSDI submission was accepted! Clockwork is a DNN-serving system that exploits the fact that deep-learning computations have extremely predictable performance. Pre-print here: https://t.co/BWDrasNyaF
Excited that our work on efficiently serving deep neural networks (DNNs) will appear at #OSDI 2020! (Spoiler: consolidate all system choices to make performance predictable)
Preprint version: https://t.co/9e7OFSSO22.
@0xREZA@mpi_jcmace@SimbiosysLab