SWu commented on issue #4519: [Relay] VM executor doesn't work with higher order gradient function URL: https://github.com/apache/incubator-tvm/issues/4519#issuecomment-566620028 I figured out the issue I was running into, which I believe is a regression in relay, so I've opened a separate issue detailing it here: https://github.com/apache/incubator-tvm/issues/4534#issue-539162046 After increasing my ulimit, the following works with VM: ```python mod = relay.Module.from_expr(relay.transform.gradient(loss_func)) seq = relay.transform.Sequential( [relay.transform.PartialEvaluate(), relay.transform.DeadCodeElimination()] ) with relay.build_config(opt_level=3): mod = seq(mod) exec = relay.create_executor(kind="vm", mod=mod) loss, grads = exec.evaluate()(label, data, *params) ``` It would be useful to add this information to the relay documentation about running higher order gradient functions in the VM
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
