null_loss问题：标签值超出了类别范围？

[GaeHyo]

请提出你的问题 Please ask your question

任务日志如下：
File "/root/paddlejob/workspace/code/main.py", line 458, in
main()
File "/root/paddlejob/workspace/code/main.py", line 150, in main
train(train_loader, model, criterion, optimizer, epoch, normalizer)
File "/root/paddlejob/workspace/code/main.py", line 206, in train
loss = criterion(output, target_var)
File "/home/aistudio/.local/lib/python3.10/site-packages/paddle/nn/layer/layers.py", line 1429, in call
return self.forward(*inputs, **kwargs)
File "/home/aistudio/.local/lib/python3.10/site-packages/paddle/nn/layer/loss.py", line 921, in forward
return F.nll_loss(
File "/home/aistudio/.local/lib/python3.10/site-packages/paddle/nn/functional/loss.py", line 1448, in nll_loss
out, total_weight = _C_ops.nll_loss(
ValueError: (InvalidArgument) label should not be out of bounds.
[Hint: Expected cur_label >= 0 && cur_label < n_classes == true, but received cur_label >= 0 && cur_label < n_classes:0 != true:1.] (at /paddle/paddle/phi/kernels/cpu/nll_loss_kernel.cc:68)

/mnt
[INFO]: size checker 352 finished

本地demo已跑通，在ai studio上报错
本地paddle版本：paddle 3.0 beta1

[GaeHyo]

请问有人负责解答吗？

[TingquanGao]

可以提供下复现命令吗？

[GaeHyo]

cgcnn-paddle.zip

本地环境：
python 3.10
paddle 3.0.0
cuda 12.3