I'm using tensorflow in python and I'm exploring the hyper parameters of a machine learning models for my dataset. My workflow involves two python scripts.
Script 1 iterates through a suite of ML hyper parameters calling a second python script that fits resulting ML model to the data.
Script 2 is command line implementation of the ML model setup and fitting to the data.
On my Macbook pro (M1 chip) everything works fine.
On my new Mac Studio, script 2 hangs after a number of calls because it fails to run on the GPU. I cannot avoid this behaviour. If I kill the hung job the first script continues to call script 2 until it hangs again after another 10ish calls of script 2.
What going on? Problem with the GPU of Mac studio?
Any suggestion to test would be appreciated.
R