mitiskuma commented on PR #18871: URL: https://github.com/apache/tvm/pull/18871#issuecomment-4000085183
I think we'd lose a 10-15% of the performances without caching, I can get bencharmarks if you think it's something to explore. meanwhile I get you few models comparison vs base: Qwen3 0.6B | base: 50t/s vs improved: 150t/s Qwen3 1.7B | base: 49t/s vs improved: 117t/s Qwen3 4B | base: 34t/s vs improved: 69t/s -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
