Where to run longformer sequence classification and best params

Hello, I am using longformer for sequence classification. My max token size is 4096 and my dataset has 276520 rows. I am trying to do training on my dataset starting with just 10,000 rows but it is taking a really really long time. I first started running it on saturn cloud on gpu but it keeps getting stuck without moving at any specific point and after 2 hours I just stop it.
I am wondering where is the best place to run my model like aws has emr or sagemaker and also, what are the best parameters to use for the amount of data that I have.
If anyone can help it would be very appreciated.