AndrewZhaoLuo opened a new pull request #9849: URL: https://github.com/apache/tvm/pull/9849
Add some unregistered ops to the default AMP lists. By default the pass places unregistered ops to the follow list so this does not change behavior. It does suppress some errors though. `nn.depth_to_space` -- this is just a fancy reshape op so should be in follow list `argmax`, `argmin` -- this one is pretty much the same computation as in `min` and `max` which are in the follow lists `fast_*` -- these fast operations use approximations involving polynomials in a clipped domain of the normal functions. They seem safe, though `fast_tanh` has coefficients outside of representable range of fp16 (which I think is fine) `clip` -- usually used to implement relu6, might have problems if clip range are outside of fp16 range but this should be rare `image.resize2d` -- The amount of accumulation is pretty minimal Excepting the `fast_*` functions which have no analog I know of, this matches closely with TFs mixed precision list (github.com/tensorflow/tensorflow/blob/v2.5.0/tensorflow/core/grappler/optimizers/auto_mixed_precision_lists.h) so feel confident about this change. -- 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]
