🆑
κ = α|working> + β|slacking>
Making Composable & Scalable Stack for Data-Intensive Applications!⚓Backend Compiler Engineer@NVIDIA⚓PhD (@illinois-impact)⚓BEng (Tsinghua)
-
Univ. of Illinois at Urbana-Champaign
- IMPACT Group, Urbana, IL
-
21:02
(UTC -06:00) - kunwu.me
- https://orcid.org/0000-0002-0149-1409
- in/kun-wu-069a14105
- https://go.kunwu.me/wakatime
Highlights
Pinned Loading
-
pytorch-direct_dgl
pytorch-direct_dgl PublicPyTorch-Direct code on top of PyTorch-1.8.0nightly (e152ca5) for Large Graph Convolutional Network Training with GPU-Oriented Data Communication Architecture (accepted by PVLDB)
-
pytorch-direct
pytorch-direct PublicCode for Large Graph Convolutional Network Training with GPU-Oriented Data Communication Architecture (accepted by PVLDB).The outdated write-up (https://arxiv.org/abs/2101.07956) explains engineeri…
-
hst10/pylog
hst10/pylog PublicPyLog: An Algorithm-Centric FPGA Programming and Synthesis Flow
-
-
mlir-standalone-template
mlir-standalone-template Public templateForked from jmgorius/mlir-standalone-template
An out-of-tree MLIR dialect template w/ CI flow to keep up-to-date.
CMake
-
CV-tsinghua-template
CV-tsinghua-template Public templateAll hail, Thy Highest University (THU)
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.