masahi commented on issue #8561: URL: https://github.com/apache/tvm/issues/8561#issuecomment-887018415
Yes, I'm aware that TF2 models have large number of input boxes and our NMS becomes a bottleneck. I think our NMS needs further improvement in general. cc @trevor-m Two questions: 1. Do you have numbers for `TF-TVM` before https://github.com/apache/tvm/pull/8174? You can disable the new conversion in 8174 by changing https://github.com/apache/tvm/blob/main/python/tvm/relay/frontend/tensorflow_ops.py#L914 to `if False:`. 2. Does `TF latency` number include CPU time? Because I heard that TF doesn't support running combined NMS on GPU. -- 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]
