Hello,
I was able to make DeepSpeed and Ray (for hyperparameter search) work on my machine separately. I was wondering if it would be possibile to make them work together at the same time. I need them to work at the same time otherwise I am not able to fine tune my model (GPT2). I tried to pass the argument --deepspeed in the config.json and then using trainer.hyperparameter.search() but with no success. I do not think it is possibile. Am I wrong? thanks