The high-level output of the VLM is really interesting for understanding the output of the flow-matching head and I think it will offer some flexibility in fine-tuning -- is there a reason it is not open-sourced yet, and is there a possibility of open-sourcing it in the near future?
The high-level output of the VLM is really interesting for understanding the output of the flow-matching head and I think it will offer some flexibility in fine-tuning -- is there a reason it is not open-sourced yet, and is there a possibility of open-sourcing it in the near future?