StarCoder: may the source be with you! — mentioned by sestinj

Wed, 26 Jul 2023 23:41:07 GMT

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), 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…

Paper and discussion history. A mention is not an endorsement.

StarCoder: may the source be with you! — HN submission

Mon, 15 May 2023 21:06:05 GMT

StarCoder and StarCoderBase: 15.5B parameter models with 8K context length

317 points; 162 discussion replies on HN. Paper and source history. A submission is not an endorsement.

StarCoder: may the source be with you! · HackerPapers discussions

StarCoder: may the source be with you! — mentioned by sestinj

StarCoder: may the source be with you! — HN submission