Code error

[lzjjia]

Hi, I followed the steps you gave to run the code, but an error was reported in line 632(feat_fc_video_source = feat_fc_source.view((-1, num_segments) + feat_fc_source.size()[-1:])) of the models.py file.
the error as follows:
Traceback (most recent call last): File "/data/lizhijia/python_project/TA3N-master/main.py", line 835, in <module> main() File "/data/lizhijia/python_project/TA3N-master/main.py", line 252, in main prec1 = validate(val_loader, model, criterion, num_class, epoch, val_file) File "/data/lizhijia/python_project/TA3N-master/main.py", line 707, in validate _, _, _, _, _, attn_val, out_val, out_val_2, pred_domain_val, feat_val = model(val_data, val_data, [0]*len(args.beta), 0, is_train=False, reverse=False) File "/home/lizhijia/anaconda3/envs/Hyrsm++/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "/home/lizhijia/anaconda3/envs/Hyrsm++/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 165, in forward return self.module(*inputs[0], **kwargs[0]) File "/home/lizhijia/anaconda3/envs/Hyrsm++/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "/data/lizhijia/python_project/TA3N-master/models.py", line 632, in forward feat_fc_video_source = feat_fc_source.view((-1, num_segments) + feat_fc_source.size()[-1:]) # reshape based on the segments (e.g. 640x512 --> 128x5x512) RuntimeError: only one dimension can be inferred

[lzjjia]

I noticed that when training to train[1][0/1], the parameters in the viem method became (-1,-1,512), so I fixed the parameters in the view method directly to (-1,5,512), and then the code can run. But the training result is:
Best score 33.33333206176758 vs current score 12.962963104248047
Is this result correct?

[lzjjia]

The source domain I used is UCF101 and the target domain is olympic

[lzjjia]

one more question: You divide the dataset into a validation set and a training set. When I test the model, should I use the validation set of the target domain?