Nlp

Notes from a few rounds of compiling Flash Attention on an A800 server. I walk through checking that ninja is actually working, limiting the build to the GPU architecture I need, and balancing parallel jobs against RAM usage. If your build takes forever or ends with a “killed” error, these are the settings and pitfalls I wish I had checked first.

Briefly record the conclusions and methods that appear in the official documentation of hugging face. By the way, record some personal thinking and confusion, so that it is convenient to go back and solve the problem later. Welcome to comment.