altanh commented on issue #8579: URL: https://github.com/apache/tvm/issues/8579#issuecomment-888827589
**Update:** the problem goes a bit deeper with the Interpreter executor. Specifically, the function returned by `intrp.evaluate(fwd_func)` is actually a Python closure which itself creates the new interpreter each time it is invoked. This means we can't even get around the caching problem by just calling `.evaluate` once. **Possible solutions:** 1. just give up on using the interpreter executor for now 2. rewrite the `evaluate` function so that it only creates 1 interpreter (and has the closure call it) rather than a new interpreter each time the closure is invoked I'll go ahead with option 2 in the hot fix PR for now -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
