lhotse: load_state_dict doesn't catch 'StopIteration' exception

Hi everyone,

I am using icefall to train a RNNT model and after I start from a batch and load the sampler sate, the dynamicbucketsampler raise an StopIteration which resulted the program to abort. I think that “exception” should have been caught somewhere up the stack trace and processed.

-- Process 0 terminated with the following error:
Traceback (most recent call last):
  File "/home/azureuser/anaconda3/envs/icefall-torch1.10cu102/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 59, in _wrap
    fn(i, *args)
  File "/home/azureuser/codebase/icefall-birch/egs/gigaspeech/ASR/pruned_transducer_stateless2/train.py", line 869, in run
    train_dl = gigaspeech.train_dataloaders(
  File "/home/azureuser/codebase/icefall-birch/egs/gigaspeech/ASR/pruned_transducer_stateless2/asr_datamodule.py", line 320, in train_dataloaders
    train_sampler.load_state_dict(sampler_state_dict)
  File "/home/azureuser/codebase/lhotse-birch/lhotse/dataset/sampling/dynamic_bucketing.py", line 174, in load_state_dict
    self._fast_forward()
  File "/home/azureuser/codebase/lhotse-birch/lhotse/dataset/sampling/dynamic_bucketing.py", line 190, in _fast_forward
    next(self)
  File "/home/azureuser/codebase/lhotse-birch/lhotse/dataset/sampling/base.py", line 261, in __next__
    batch = self._next_batch()
  File "/home/azureuser/codebase/lhotse-birch/lhotse/dataset/sampling/dynamic_bucketing.py", line 233, in _next_batch
    batch = next(self.cuts_iter)
StopIteration