Will there be training based on qwen2.5 later
Will there be training based on qwen2.5 later? Does it support video input and enable analysis of spatial positions and path planning in videos?
I hope they release the finetuning script in a way that we can train it on our own with any LLM model
We will release our training code so other people can experiment with this kind of thing. While we do plan on releasing future versions with improvements, which could include video or new LLMs, what exactly those are has not yet been decided.
@chrisc36 Can you please release the datasets?
@SinanAkkoyun We are trying to release the PixMo dataset by the end of November at the latest. For now, you can sign up to receive updates using this form: https://docs.google.com/forms/d/e/1FAIpQLSdML1MhNNBDsCHpgWG65Oydg2SjZzVasyqlP08nBrWjZp_c7A/viewform?usp=sf_link
@amanrangapur Thank you! ☺️