As many of you know, we recently added some capabilities in TVM to interact with hardware implementations (RTL) for improving design productivity, see this blogpost and PR for more information. There are plenty of opportunities next from: running an actual ML workload, adding support for other hardware simulators (i.e., commercial), evaluating more complex hardware designs, among others. We would like to get an idea from the community on how they could benefit from these new features in TVM? and also see what capabilities people are most interested in based on their use cases?
5 Likes
We’d love to get input/feedback suggestions etc. Thanks!