Anglin Liu
Anglin Liu
> I guess you can only fine-tune the detector and the shared backbone. > > According to the current implementation, it appears that training code for the tracker modules is...
> How much compute do you need to fine-tune the model? How can we fine-tune it without updating all 840M parameters? When the batch size is 1 and the resolution...
> [@Joey-S-Liu](https://github.com/Joey-S-Liu) Did you fine-tune them only using bbox? In my case, when I turn on the enable_segmentation, it said OOM error even if I used H200 and 1 batch...
> > [@Joey-S-Liu](https://github.com/Joey-S-Liu) Did you fine-tune them only using bbox? In my case, when I turn on the enable_segmentation, it said OOM error even if I used H200 and 1...
> > > [@Joey-S-Liu](https://github.com/Joey-S-Liu) Did you fine-tune them only using bbox? In my case, when I turn on the enable_segmentation, it said OOM error even if I used H200 and...
Hi, have you solved this problem?