Luo Cong

> Hi @nullkatar, I am working on a customized dataset but facing problems, could you also send me an email with any steps that might help. My email is [email protected]....

The new url is https://homes.cs.washington.edu/~ranjay/visualgenome/ (I find it at [this website](https://huggingface.co/datasets/visual_genome/discussions/3)) I think you can also load the dataset directly from huggingface dataset: https://huggingface.co/datasets/visual_genome

@Ming-er I think the only difference between 1x1 conv and linear is that conv layer ignores the bias term in linear layer, and except that the calculation is the same....

https://github.com/Zejun-Yang/AniPortrait/blob/cb86caa741d6ab1e119ea7ac2554eb28aabc631b/train_stage_1.py#L166-L169 @Mxwgreat `random.randint(0, dataset_len)` can generate random integer **including** `dataset_len`, and `dataset_len` will cause index out of range error because the index range for `valid_dataset` is 0~`dataset_len-1` and you can...

Is the ann_file `data_pairnet.json` same as the `vg150.json` at this [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EiBEV1Z3ueBJqJVO4j7z0YwBt_Jvj2AqYTRsiIs-8pZowg?e=C2O5yg), which I find in [OpenPSG](https://github.com/Jingkang50/OpenPSG) repo.

any update on this question? I'm also little confused and can't find related literature about this strategy.

@XiaoluDing Have you figured it out? I'm also a little confused.

@JZArray Hi, I want to ask do you know how to train on images? the paper says they train the model on both images and videos, but image just has...

Thanks for the suggestions! I'll give them a try. I've noticed that the official AnimateDiff code doesn't use gradient checkpointing by default, and it can save lots of GPU memory.

@massyzs I can't find either. Seems that style adaptation and scale adaptation are not implemented in the code?