Comment by 0xdeadbeefbabe

I ask the LLM to generate the code but mask the last token. Then I do softmax and give it an answer. Pretty soon my perceptrons are more connected than ever.