Will mm-video dataset be released?

[kon-pap]

Thank you for releasing traces for multimodal inference, it has been very helpful!

I wanted to ask whether you are also planning to release the mm-video or mm-omni datasets used in the paper. Having access to video traces (at least number of video tokens and request timestamps) would be extremely valuable for extending research on multimodal tasks.

If so, could you share whether there is a timeline or plan for making it available?

Thanks again for your great work!

[kon-pap]

@lixueclaire any update on this issue?

[lixueclaire]

Thank you for releasing traces for multimodal inference, it has been very helpful!

I wanted to ask whether you are also planning to release the mm-video or mm-omni datasets used in the paper. Having access to video traces (at least number of video tokens and request timestamps) would be extremely valuable for extending research on multimodal tasks.

If so, could you share whether there is a timeline or plan for making it available?

Thanks again for your great work!

Thank you for your interest in our work! I apologize for the delayed response.

We initially released only the image data because it had the largest sample size, while the samples for video are relatively limited. However, with the recent increase in usage, we do plan to release the mm-video and mm-omni datasets once we have collected more samples. We appreciate your understanding and support.