When I set num_works> 0, there is a error Producer process has been terminated before all shared CUDA tensors released

[Struggle-Forever]

[akihironitta]

lr_lambda needs to be picklable in your code (or in the library you use) in some strategy:

• https://pytorch-lightning.readthedocs.io/en/1.7.7/accelerators/gpu_faq.html?highlight=pickle#i-m-getting-errors-related-to-pickling-what-do-i-do
• https://docs.python.org/3/library/pickle.html#what-can-be-pickled-and-unpickled

[Struggle-Forever]

Thank you! I love pytorch-lightning!

[danielajisafe]

In my case, this issue was because my tensors were already placed in cuda within the getitem() and I also had torch.set_default_tensor_type('torch.cuda.FloatTensor') before the general main function in the code which technically was calling cuda on cuda tensor. So I either dropped one of them or put everything on CPU as my task did not require a GPU.