LeeSureman / Batch_Parallel_LatticeLSTM

Chinese NER using Lattice LSTM. Reproduction for ACL 2018 paper.
130 stars 16 forks source link

How to deal with the ValueError:Expect torch.optim.lr_scheduler for LRScheduler. Got <class 'torch.optim.lr_scheduler.LambdaLR'> #32

Open cflyee opened 2 weeks ago

cflyee commented 2 weeks ago

如题,使用torch2.0.0,cuda为1.8,但torch版本可能过高导致传入的LRSchduler变成了LambdaLR,这种情况如何解决呢,如果降低torch版本又对应不上cuda......