Jack Lanchantin

Can you try running without a GPU? On Wed, Feb 15, 2023 at 3:30 AM ans92 ***@***.***> wrote: > Yes I were able to download and run the model on...

It could be. I can't replicate the error

> Hi, Thank you for the code and the great work. I have two questions regarding your work. First how many number of epochs your model needs to train successfully?...

Are you able to download the example datasets (coco, voc)? If so, you should be able to simply match the format of those datasets.

I've updated the urls. You should at least be able to download the original datasets.

Hi, this can be visualized using the `attn_output_weights` that are returned from `nn.MultiheadAttention`: ``` multihead_attn = nn.MultiheadAttention(embed_dim, num_heads) attn_output, attn_output_weights = multihead_attn(query, key, value) ```

the adjacency matrix processing is done here: https://github.com/QData/LaMP/blob/master/utils/utils.py#L86 does that help?

That's the train split adjacency matrix (should be either a full adjacency matrix or sparse representation).

There are 1,141 samples (see table 5 in paper) See the `adj_insts` var in DataLoader. That's what sider uses