Several open-source kernels in this repository originated in NVIDIA's CuTe DSL kernel library. We gratefully acknowledge the contributors who helped develop, bring up, and test these kernels.
Contributors include:
- Rachit Garg rachitg@nvidia.com
- Jack Yang jackyang@nvidia.com
- Hao Sheng hsheng@nvidia.com
- Alex Li alel@nvidia.com
- Aragorn Guan aragorng@nvidia.com
- Bangyu Shen bangyus@nvidia.com
- Caleb Du cadu@nvidia.com
- Xiao Song xiaos@nvidia.com
- Siddhartha Raman sraman@nvidia.com
This acknowledgement covers the CuTe DSL kernel work now represented by these
modules in python/cudnn/:
gemm/cutedsl/dense/amaxgemm/cutedsl/dense/dsrelugemm/cutedsl/dense/srelugemm/cutedsl/dense/swiglugemm/cutedsl/grouped/dglugemm/cutedsl/grouped/dsrelugemm/cutedsl/grouped/dswiglugemm/cutedsl/grouped/glugemm/cutedsl/grouped/glu_hadamardgemm/cutedsl/grouped/quantgemm/cutedsl/grouped/srelugemm/cutedsl/grouped/swiglugemm/cutedsl/grouped/wgradgemm/cutedsl/discrete_grouped/dswiglugemm/cutedsl/discrete_grouped/swiglurmsnorm_rht_amax
Thank you also to the broader CUTLASS/CuTe DSL and infrastructure teams who supported the original kernel development.