Important things I forgot to mention in the original tweet!

- SQLCoder is fine-tuned on StarCoder, an awesome initiative of the BigCode

- We used a slightly novel training approach. We first trained the model on 'easy' questions, and then trained the result of that on…

We just open-sourced SQL Coder, a 15B param text-to-SQL model that outperforms OpenAI's gpt-3.5! When fine-tuned on an individual schema, it outperforms gpt-4. github.com/defog-ai/sqlco…

The model is small enough to run on a single A100 40GB in 16 bit floats, or on a single…

Other interesting things

- BigCode and Argilla have the most downloaded datasets
- Open Assistant and Together AI are among the orgs with the most liked datasets
- Sentence Transformers keeps up with the most downloads

LLMs are great at programming tasks... for Python and other very popular PLs. But, they are often unimpressive at artisanal PLs, like OCaml or Racket. We've come up with a way to significantly boost LLM performance of on low-resource languages. If you care about them, read on!

In the past two weeks, we've seen 4 new code models drop: StableCode, OctoCoder, OctoGeeX, and DeciCoder 🚀

Everyone's talking about 'HumanEval' – so how does code evaluation work & what makes reproducibility challenging?

A thread 🧵:

Next Tuesday, I will give a webinar hosted by Analytics Vidhya on the training of LLMs for code, like StarCoder.

I will also discuss how to leverage these models using open-source libraries such as transformers, datasets and PEFT.

Register here: bit.ly/47BSnqK.

How to instruction tune Code LLMs w/o #GPT4 data? Releasing

🐙🤖OctoCoder & OctoGeeX: 46.2 on HumanEval🌟SoTA🌟of commercial LLMs
🐙📚CommitPack: 4TB of Git Commits
🐙🎒HumanEvalPack: HumanEval extended to 3 tasks & 6 lang

📜arxiv.org/abs/2308.07124
💻github.com/bigcode-projec…
1/9

For all PhD students in small labs: find all possible ways to collaborate with well-known open research groups like EleutherAI LAION BigScience Research Workshop BigCode; apply to every single fellowship and look for connections. It’s not optional if you want to have a career.

The power of building tools, datasets, and models in the open: the community can build on top of it and everyone profits!

Exhibit A: since the release of 📑The Stack and ⭐️StarCoder research groups from academia and industry have trained models on top BigCode's releases.

Inspired by the Open LLM LeaderBoard, and with several strong code models released, we created a Multilingual Code Leaderboard:

📊 10+ programming languages
⚡Throughput measurement
🔬 Fully reproducible
✉️ Open for submission of results

huggingface.co/spaces/bigcode…

WizardCoder attains the 🏆1st position on BigCode Multilingual Code Models Evaluation Leaderboard, which compares performance of base multilingual code generation models on HumanEval benchmark and MultiPL-E.

👉huggingface.co/spaces/bigcode…

Please join us in congratulating the BigCode community for the latest batch of 1B, 3B, 7B open-access StarCoder models for code-generation (and more), released today under a commercial use OpenRAIL-M license. Incredible work!

Join the collaboration here: bigcode-project.org/docs/about/joi…

We got a code model leaderboard, courtesy of BigCode 🔥🔥🔥 huggingface.co/spaces/bigcode…

Sean Hughes from ServiceNow Research and Loubna Ben Allal from Hugging Face talk about their joint BigCode and how Toloka contributed to its success. Watch now: youtu.be/a_d36OBK5Lk

AI code completion running 100% locally inside your browser, thanks to BigCode's StarCoder models and 🤗 Transformers.js!

We also got their new 1B model running at ~20 tokens per second in Node.js (CPU). 🚀

Check out the demo! huggingface.co/spaces/Xenova/…

Starcoder 3B runs on CPU ⚡

Excited to launch Hugging Face's Starcoder model in gpt4all.io on CPU!

Local code models will be everywhere