Have you thought about generating the SFT data for Deepseek-Coder-V2?

[tangzhy]

Given its exceptional capabilities in coding and mathematics, the accuracy of both can be automatically verified by the final results. It would be quite persuasive if your method could match their reported performance.

[zhangchen-xu]

Hi, Thanks for the suggestion! I already put the run scripts of Deepseek-Coder-V2 here. It can work! However, most of our computing resources focus on the Gemma 2 series, which has no copyright claim. So we don't start extracting Deepseek data for now.

[zhangchen-xu]

We also mentioned the plans for extracting Deepseek-Coder-V2 in our internal meeting. Once Gemma 2 is done, we may come back!