Hi! It seems a lot of people are experiencing issues with the expected time it takes to train a model, including myself. You just explained that we need a fused version of models/ops for the GPU acceleration (please correct me if I am wrong or interpreting it wrong). Would it therefore maybe a good idea to publish some of of list of the fused models/ops so we know which ones are GPU acceleration supported or not? Then we all know what we can expect from which models and you won't get dissatisfied users that post to this to developer forum. Thanks in advance for any clarification!

This also didn't work for me as tensorflow-macos seems does not seem to work with any numpy version above 1.23*: ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. tensorflow-macos 2.12.0 requires numpy<1.24,>=1.22, but you have numpy 1.24.3 which is incompatible. However, fortunately the solution from @alexander_the_coder did work for me!