26 ms·

What do you think about starcoder

• sestinj 3y ago
  We've experimented with it quite a bit, and at one point used it as our default model for inline edits. As they discuss in section 5.1 of the paper (https://arxiv.org/pdf/2305.06161.pdf https://arxiv.org/pdf/2305.06161.pdf), they've trained on data formatted as "<commit_before>code<commit_msg>text<commit_after>code<eos>", which was convenient for our use case. Unfortunately, it tended to repeat itself and for more complex edits it doesn't match the raw ability of gpt4. I'm optimistic about open-source models though, and the data Continue lets users collect for themselves will help the open-source community get the data they need to compete.