Also, I was also experiencing memory leaks w/ tensorflow-metal when doing large hypertuning runs, and originally thought it must have been related to my model, but since switching to CPU haven't experienced the same issues.

It sounds like the issue is related to what is posted on this page: https://developer.apple.com/metal/tensorflow-pl CPU performance is faster than GPU on your network. Find out if your workload is sufficient to take advantage of the GPU. On small networks running with small batch sizes, the CPU may perform faster overall due to the overhead related to dispatching computations to the GPU. This will get amortized when the batch or model sizes grow, since the GPU can then take better advantage of the parallelism in performing the computations.

If I'm not mistaken, I think this issue is more apparent on Sonoma than Ventura.

Out of curiosity, is your M1 Ultra running Sonoma or Ventura?

I believe it is due to process deadlocking, if I'm not mistaken. Some people are skipping the GPU altogether on Mac Silicon.

I've experienced the issue where streams aren't starting w/ Safari 15x (both Mac and iOS), and discovered that it seems to fix itself (in my project, at least) if stopping the outgoing stream (per audio / video kind), then restarting the stream (i.e. making a new gUM [getUserMedia] call). The "restart" method I just mentioned seems to also work if choosing an entirely different capture device, of the same audio / video kind, on the second attempt. i.e. first microphone capture fails for one device, starting another device works. Even stranger, for me, is my project is based in React, and I've worked around this w/o doing the restart method in development mode, but in production, it is persistent, and for the life of me I can't figure out what's changed.