I was just wondering, what is the best usage for the attn_debug flag. It seems to only print attention scores to the terminal in a kind of difficult format, is there some standard way to save the scores in a standard format?
I was just wondering, what is the best usage for the attn_debug flag. It seems to only print attention scores to the terminal in a kind of difficult format, is there some standard way to save the scores in a standard format?