Linjie Li

> Dear scholar, This is your code in pos_emb.py > > ``` > y_diff = center_y[i] - center_y[j] > x_diff = center_x[i] - center_x[j] > diag = math.sqrt((y_diff)**2 + (x_diff)**2)...

>  >  >  > 0: wearing, > 1: holding, > 2: sitting on, > 3: standing on, > 4: riding, > 5:eating, > 6:hanging from, > 7:carrying,...

Hi there, Thanks for your interests in our project. The eval.py did not include the weighted sum part. As shown in equation 10 of the paper, the weighted sum is...

We don't have a decoder in UNITER Model, which was also not in our paper. I am not sure which paper you are referring to?

`train_datasets` is indeed in opts (which is loaded from config) https://github.com/ChenRocks/UNITER/blob/1dbfa62bb8ddb1f2b11d20775ce324c5e45a3ab4/config/pretrain-vcr-base-4gpu.json#L30

Just verified from my end that clipbert_image_text_pretrained.pt/bert-base-uncased.tar/grid_feat_R-50.pth can be downloaded without error. I would suggest to run bash scripts/download_pretrained.sh again.

If you run `source tools/download.sh`, you can find pretrained checkpoints downloaded under `pretrained_models`.