Zhao Zhang
Zhao Zhang
> @lecoan hello, the problem still exists when I copy citations. I have installed PyObjC with `pip install PyObjC`, and here is the debug log: > > ``` > [10:15:05.297]...
Does it support direct calculating loss using mini batch? like: -ssim_loss(ims, labels), where ims and labels are both [bs, 3, h, w]
Please!
I found that if I load the pre-trained checkpoint, save it, and then load it (there is no training during this period). The results will be very different. I don't...
> I also came across such problem (the zero-shot performance of the official checkpoint drops if we load and save it (without training)). I try to compare the hyper-parameters in...
Some detailed differents between resaved ckpt and [ofa-base ckpt](https://github.com/OFA-Sys/OFA/blob/main/checkpoints.md) are below. 1. Our training task removes MLM and images infilling. (From the paper tab.10, it should not be the main...
> @zzhanghub @Charles-Xie Changing the dimension of `image_position_idx` to 1026 is to be consistent with the dimension of `embed_positions`, it shouldn't affect the performance. I guess you save the model...
> @zzhanghub There could be many reasons: > > 1. We use Pillow to open and save the picture, and the picture information will be lost in this process. I...
同样的问题，很困扰