Tong

What is the definition of **targeted attack**? For graph data, the target can represent two things: - target node: attack specific node (in node classification task) - target label: attack...

In the notebook, you choose [9,11,15,16,17] as the swapping layer. I wonder about the consideration of this choice. Thanks in advance.

For current `visualize_graph`, the node colors are default 'white', which is not as intuitive as in GNNExplainer: target node is red, different node class with different node color. We add...

Re-Organize explanation examples, more effective evaluation of explanations

5

### 🚀 The feature, motivation and pitch Explanation for graph data is not as intuitive as images, therefore, proper evaluation is very important. The current explanation examples are not well-organized,...

Code for initialize CausalConv3d from pretrained Conv2D.

10

Hi, I find that you introduce in CausalVideoVAE.md that you use special initialization(tail initialization) for CausalConv3d training. I am interested in this trick, and I would be sincerely grateful if...

The performance gap between using vanilla 3dVAE and Casual3dCNN

1

Hi, I am interested in the performance of using 3d CNN without casual for VAE training, and corresponding results of generation. As I understand, the propose of Casual3dCNN is for...

I have a question on the feature sets. As the dimensions of c1, c3, c7 are very large, how did you represent each protein using 50-dimensional vectors? Thanks in advance!...

Hi, is there any mapping files about correspondence between protein name and ID in dataset?

Must initialize from damo-vilab/text-to-video-ms-1.7b?

10

Hi, I found that here: https://github.com/Picsart-AI-Research/StreamingT2V/blame/c1b8068bcbcdbbfa0dd0df3371d3c93a1f5132de/t2v_enhanced/model_init.py#L71C7-L71C7 ``` def init_streamingt2v_model(ckpt_file, result_fol): ... cli = CustomCLI(VideoLDM) ``` It seems that VideoLDM must initialized from damo-vilab/text-to-video-ms-1.7b (in config pipeline_repo: damo-vilab/text-to-video-ms-1.7b). Moreover, ckpt_file, which...

How to guarantee the relaxation from {0, 1} to [0, 1] results in optimized s contains values close to 0 or 1, but not 0.5?

3

I wonder if the relaxation from $s\in\{0,1\}$ to $s\in[0,1]$ can guarantee the optimized $s$ contains values close to 0 or 1. If most of the values are like $0.45, 0.55$,...