Apply for community grant: Personal project (gpu)

#1
by shauray - opened

StarCoder is not the top-of-the-line model you can use to code BUT it has the lowest inference cost-to-performance ratio in my opinion it does work with the standard CPU but -

  1. not enough tokens
  2. takes a million years to output something
    With a GPU grant I can improve the inference time and more people can use it, it's 1B with some good good results and to prove it here's a key result from the author itself [39 @1pass human eval]
shauray pinned discussion

Sign up or log in to comment