RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation · HackerPapers discussions

RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation — mentioned by synthiq

Sat, 03 Jun 2023 16:39:45 GMT

For anyone interested in related research, I used https://mirrorthink.ai to find some background on the state-of-the-art. (disclaimer: this is AI generated, but grounded on contents of papers, with real references, so I'd say it is still constructive) The state-of-the-art in code generation has seen significant advancements with the deployment of large language models (LLMs) in various code authoring tools. One such example is the study on GitHub Copilot, Amazon CodeWhisperer, and ChatGPT [1], which evaluates the code quality of these AI-assisted code generation tools. The study reveals that…

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

RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation — mentioned by wsxiaoys

Thu, 06 Apr 2023 17:51:37 GMT

Tabby's philosophy is to achieve a completion rate comparable to Codex/Copilot by using a model size of less than 1B, with support for BF16/FP16 that reduces VRAM requirements to 2G or less. This may seem impossible, given that the model size is 10 times smaller than that of Codex, but it is definitely achievable, especially in an on-premises environment where customers want to keep the code behind a firewall. Related research works include [1]. (Hints: combine code search and LLM). [1]: RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation…

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

RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation — mentioned by wsxiaoys

Wed, 29 Mar 2023 00:49:01 GMT

That‘ll be super interesting! Local context is definitely the key factor for small models achieving better quality than copilot (related: [1], [2]) . One things I’d really wanna have in Sourcegraph: A Search API supports custom retrieval / ranking. Research works (e.g [2]) show simple BoW fetched context is more efficient for code completion tasks. Disclaimer: I’m building https://github.com/TabbyML/tabby an open source alternative of copilot. [1]: https://arxiv.org/abs/2206.12839 [2]: https://arxiv.org/abs/2303.12570

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

RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation — mentioned by wsxiaoys

Tue, 28 Mar 2023 15:39:51 GMT

Cody is more like a wrapper on top of GPT API (gpt-3.5-turbo) endpoint, with related code snippets embed in prompt. It’s not competing the low latency scenario that Github Copilot seats in. There’re recent works show a potential of beating Copilot’s performance (e.g https://arxiv.org/abs/2303.12570) with much smaller models (500M vs 10B+). Inspired by these work, I’m building Tabby (https://github.com/TabbyML/tabby), a OSS GitHub Copilot alternative. Hopefully it could make low cost AI coding accessible to everyone.

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

RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation — HN submission

Sat, 25 Mar 2023 00:40:23 GMT

RepoCoder: Repo-Level Code Completion Through Iterative Retrieval and Generation

4 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.