GTC'22 Spring Cutlass Talks

[hwu36]

There are four Cutlass talks in the coming GTC

• Accelerating Backward Data Gradient by Increasing Tensor Core Utilization in CUTLASS [S41996]: @manishucsd will introduce our super fast strided dgrad which can also be used as dconv as shown in Transposed conv2d and wgrad split k examples #413. He will also give a general CUTLASS update. If you use CUTLASS heavily, don't miss it.
• Connect with the Experts: NVIDIA Math Libraries [CWE41721], @kerrmudgeon , @manishucsd , and @hwu36 will do live Q&A about CUTLASS. Last year, we answered many questions from many users in a similar event.
• Use CUTLASS to Fuse Multiple GEMMs to Extreme Performance [S41606]
• Auto48: A General Framework for Automatic Model Compression and Acceleration using Int4/Int8 Mixed Precision [S41611]
• Large Models Are Not Always Expensive: Large Scale Mixture Of Expert Models With Efficient Inference Empowers MicroSoft Translator With Best Models [S42518]

[masahi]

@hwu36 The link in Accelerating Backward Data Gradient by Increasing Tensor Core Utilization in CUTLASS above is pointing to a different talk.

Just watched three presentations related to the Hopper arch. Can't wait to see how cutlass would utilize all the cool new features!!