New goal for this year! My goal is to review a paper every fortnight and put up a blog post with the review summary. I hope to definitely read more papers that those, but writing a summary that makes sense when I read it again, say 6 months later, and is able to explain the paper from scratch, is kind of a time consuming process. So the paper review series is only for those papers who I felt were interesting and that could hopefully be useful later.
My usual style of reviewing papers is quick bullets and notes of the most important points which sometimes don’t even make sense to me after a while. Hopefully, this review series will improve my writing skills as well.
The papers will all (most probably) be in the space of systems for LLMs which includes cluster management for training, inference server optimizations, MoE etc. and anything remotely related to GPUs.
I am making this a separate blog post as firstly, a whip to ensure I stick to my goal and secondly, a list to keep track of the review posts.