vacu9708 commented on issue #17977: URL: https://github.com/apache/tvm/issues/17977#issuecomment-2894702450
Hi, @Cookiee235 I don't think what you’re seeing is a bug, but rather a natural consequence of optimizing operations with float16 values. I ran your code and saw that all precision losses are within 1 ULP, except for five or fewer cases where the loss is 2 ULPs, which is still an extremely small difference. For your reference, ULP (Unit in the Last Place) means the smallest gap between representable numbers. Hope this helps. -- 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]
