Version of Pytorch and Cuda

[yuxiangwei0808]

Hi, I am repeating the experiment in the "osdi21-artifact" branch. However, I have encountered multiple jobs failure due to some errors:

Traceback (most recent call last):
  File "run_glue.py", line 750, in <module>
    main()
  File "run_glue.py", line 476, in main
    model = adaptdl.torch.AdaptiveDataParallel(model, optimizer, lr_scheduler)
  File "/root/adaptdl/adaptdl/torch/parallel.py", line 68, in __init__
    adaptdl.checkpoint.load_state(self._state)
  File "/root/adaptdl/adaptdl/checkpoint.py", line 137, in load_state
    state.load(f)
  File "/root/adaptdl/adaptdl/torch/parallel.py", line 194, in load
    state_dicts, self.gain = torch.load(fileobj)
  File "/opt/conda/lib/python3.8/site-packages/torch/serialization.py", line 600, in load
    with _open_zipfile_reader(opened_file) as opened_zipfile:
  File "/opt/conda/lib/python3.8/site-packages/torch/serialization.py", line 242, in __init__
    super(_open_zipfile_reader, self).__init__(torch._C.PyTorchFileReader(name_or_buffer))
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

which usually happens when rescaling. I think this possibly resulted from the conflict of environment. Therefore, could you please provide the versions of Pytorch, Cuda, Python, and other necessary modules?