RuntimeError: 1D target tensor expects 2D input tensors, but found inputs with sizes [46, 8, 50] and [46, 8, 50].
Closed this issue · 5 comments
Hello，when I review the process of t
![Uploading NQN42)~KA9S8C_UZTB(CC]A.png…]()
rain，this problem has been generated。 Could you tell me how to solve this question？
Hello，when I review the process of t ![Uploading NQN42)~KA9S8C_UZTB(CC]A.png…]() rain，this problem has been generated。 Could you tell me how to solve this question？
Hello, have you solved it? I also encountered the same problem
Are there details of the errors reported? I haven't encountered this problem yet
Are there details of the errors reported? I haven't encountered this problem yet
File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/train.py", line 6, in
DMD_run(model_name='dmd', dataset_name='mosi', is_tune=False, seeds=[1111], model_save_dir="./pt",
File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/run.py", line 94, in DMD_run
result = _run(args, num_workers, is_tune)
File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/run.py", line 179, in _run
epoch_results = trainer.do_train(model, dataloader, return_epoch_results=from_sena)
File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/trains/singleTask/DMD.py", line 141, in do_train
cosine_similarity_s_c_l = self.cosine(output['s_l'], output['c_l'],
File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/modules/loss.py", line 1276, in forward
return F.cosine_embedding_loss(input1, input2, target, margin=self.margin, reduction=self.reduction)
File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/functional.py", line 3490, in cosine_embedding_loss
return torch.cosine_embedding_loss(input1, input2, target, margin, reduction_enum)
RuntimeError: 1D target tensor expects 2D input tensors, but found inputs with sizes [46, 16, 50] and [46, 16, 50]
sorry，I haven‘t solve this problem，but adjusted the 2D to 3D could get the another result.Good luck! 我往北追_? @.***
…
------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年7月11日(星期二) 下午3:24 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [mdswyz/DMD] RuntimeError: 1D target tensor expects 2D input tensors, but found inputs with sizes [46, 8, 50] and [46, 8, 50]. (Issue #5) Are there details of the errors reported? I haven't encountered this problem yet File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/train.py", line 6, in DMD_run(model_name='dmd', dataset_name='mosi', is_tune=False, seeds=[1111], model_save_dir="./pt", File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/run.py", line 94, in DMD_run result = _run(args, num_workers, is_tune) File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/run.py", line 179, in _run epoch_results = trainer.do_train(model, dataloader, return_epoch_results=from_sena) File "/media/data2/zhukang/bigmodel_shared_and_private/DMD-main/trains/singleTask/DMD.py", line 141, in do_train cosine_similarity_s_c_l = self.cosine(output['s_l'], output['c_l'], File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1190, in _call_impl return forward_call(input, kwargs) File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/modules/loss.py", line 1276, in forward return F.cosine_embedding_loss(input1, input2, target, margin=self.margin, reduction=self.reduction) File "/media/data2/zhukang/anaconda3/envs/zk/lib/python3.8/site-packages/torch/nn/functional.py", line 3490, in cosine_embedding_loss return torch.cosine_embedding_loss(input1, input2, target, margin, reduction_enum) RuntimeError: 1D target tensor expects 2D input tensors, but found inputs with sizes [46, 16, 50] and [46, 16, 50] — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>
You can change to output['s_l'].transpose(0,1).contigouuous.view(labels.size(0),-1). Other outputs that use to calculate cosine_similarity can be changed in the same way. If you use pytorch1.9.1 as we used, this error will not appear.