Fail to train on gen2 dataset

[Guackkk]

Hi ! I wanted to reproduce the results in espaloma and followed the training method suggested here in https://espaloma.wangyq.net/experiments/qm_fitting.html. But the error occur when I wanted to calculate the training set performance :

Traceback ( most recent call last): File "train_gen2.py", line 86, in <module> u = torch.cat(u, dim=0) RuntimeError: Sizes of tensors must match except in dimension8. Expected size 42 but got size 21 for tensor number 1 in the list

I used the same code for training set performance and validation set performance as the one in https://espaloma.wangyq.net/experiments/qm_fitting.html : `with torch.no_grad(): for idx_epoch in range(10000): espaloma_model.load_state_dict( torch.load("%s.th" % idx_epoch) )

    # training set performance
    u = []
    u_ref = []
    for g in ds_tr:
        if torch.cuda.is_available():
            g.heterograph = g.heterograph.to("cuda:0")
        espaloma_model(g.heterograph)
        u.append(g.nodes['g'].data['u'])
        u_ref.append(g.nodes['g'])
    u = torch.cat(u, dim=0)
    u_ref = torch.cat(u_ref, dim=0)
    loss_tr.append(inspect_metric(u, u_ref))

    # validation set performance
    u = []
    u_ref = []
    for g in ds_vl:
        if torch.cuda.is_available():
            g.heterograph = g.heterograph.to("cuda:0")
        espaloma_model(g.heterograph)
        u.append(g.nodes['g'].data['u'])
        u_ref.append(g.nodes['g'])
    u = torch.cat(u, dim=0)
    u_ref = torch.cat(u_ref, dim=0)
    loss_vl.append(inspect_metric(u, u_ref))`

Thanks for any help given.

[mikemhenry]

We have made some API changes lately, I will defer to @yuanqing-wang and @kntkb to take a look at this question. Thanks!