merrymercy edited a comment on issue #6096: URL: https://github.com/apache/incubator-tvm/issues/6096#issuecomment-663218666
I ran the code and found adding `compile_engine.get().clear()` before `with autotvm.apply_history_best(log_filename):` acutally fixes the issue. This issue is not related to autotvm at all. The autotvm feature extraction is not stateful. Instead, the relay compiler is stateful. We have to clear the relay cache before using a different schedule for the same relay subgraph. @tkonolige @jwfromm Could you double-check. This issue disappears after ---------------------------------------------------------------- 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: us...@infra.apache.org
