Junyang Lin
Junyang Lin
Seemingly there are still some problems. I guess I'll provide colab notebooks for you to check it out. Perhaps after a week or so...
Wait.. I think I did not opensourced the huge for RefCOCO, so which one are you using? the pretrained one?
Did you try with the provided fairseq generator, as demonstrated in our example code?
I think it is because of the recent merges in that branch and there might still be some problems. We are now testing them and will give a feedback soon.
It should be fine to share the weights. I'll inform you after uploading.
Have you tried it on any other devices?
Any more details?
How do you evaluate overfitting? Is it about its performance biased to e-commerce or something else?
Check our README about how to finetune. We provide tutorials about how to process your own data.
Not done yet, but possible. We still need to figure out if we need to make changes on pretraining, or simply adapt the pretrained models to this task. The simplest...