With SB3, you can start training something immediately using their out-of-the-box algorithms. You can, of course, implement an algorithm yourself and have full control of the logic.
some people learn more from rolling their own, some just get lost. those who get lost should start with a framework and when the fundamentals "stick" - start implementing more of the base pieces of your program, classic or ML
Thank you so much, that’s really useful! Was always wondering how to choose between them. Good luck to your channel!
Very helpful video. Is there any reason to use sb3 rather than writing algorithms on your own using pytorch/keras/tf?
With SB3, you can start training something immediately using their out-of-the-box algorithms. You can, of course, implement an algorithm yourself and have full control of the logic.
@@johnnycode Oh that makes sense thanks.
some people learn more from rolling their own, some just get lost. those who get lost should start with a framework and when the fundamentals "stick" - start implementing more of the base pieces of your program, classic or ML