I would like to know if there is a way to deploy the TVM module to a cluster of hardware devices for performance gain. Example multiple CPUs/GPUs. It can be a homogeneous cluster.
Hi I am new to TVM and looking for similar thing. Was wondering if you got your answer. I’m looking for utilizing TVM scheduler for I/O intensive nested loops such as pattern matching and counting across collection of log files and databases. In particular I wonder if anyone has looked into combining capabilities of for cluster management https://www.ray.io/ with TVM scheduling and finer grain operation scheduling.