[BUG]: Failures when torch.distributed is not available

[amaissen]

Describe the bug

TensorDict does not work when torch.distributed is not available.

To Reproduce

Take a machine where torch.distributed.is_available() equals to False. Then import TensorDict

from tensordict import TensorDict

then it gives the following traceback:

Traceback (most recent call last):
AttributeError                            Traceback (most recent call last)
Cell In[2], line 1
----> 1 from tensordict import TensorDict
      2 import torch
      3 a = torch.rand(3, 4)

File /opt/homebrew/anaconda3/envs/aaad-poc/lib/python3.9/site-packages/tensordict/__init__.py:7
      1 # Copyright (c) Meta Platforms, Inc. and affiliates.
      2 #
      3 # This source code is licensed under the MIT license found in the
      4 # LICENSE file in the root directory of this source tree.
      6 from tensordict.memmap import MemmapTensor, set_transfer_ownership
----> 7 from tensordict.persistent import PersistentTensorDict
      8 from tensordict.tensorclass import is_tensorclass, tensorclass
      9 from tensordict.tensordict import (
     10     is_batchedtensor,
     11     is_memmap,
   (...)
     20     TensorDictBase,
     21 )

File /opt/homebrew/anaconda3/envs/aaad-poc/lib/python3.9/site-packages/tensordict/persistent.py:27
     24 import torch
     26 from tensordict import MemmapTensor
---> 27 from tensordict.tensordict import (
     28     _TensorDictKeysView,
     29     CompatibleType,
     30     is_tensor_collection,
     31     NO_DEFAULT,
     32     TensorDict,
     33     TensorDictBase,
     34 )
     35 from tensordict.utils import (
     36     _shape,
     37     DeviceType,
   (...)
     41     NUMPY_TO_TORCH_DTYPE_DICT,
     42 )
     45 class _Visitor:

File /opt/homebrew/anaconda3/envs/aaad-poc/lib/python3.9/site-packages/tensordict/tensordict.py:317
    312         return fn(*args, **kwargs)
    314     return wrapper
--> 317 class TensorDictBase(MutableMapping):
    318     """TensorDictBase is an abstract parent class for TensorDicts, a torch.Tensor data container."""
    320     LOCK_ERROR = (
    321         "Cannot modify locked TensorDict. For in-place modification, consider "
    322         "using the `set_()` method and make sure the key is present."
    323     )

File /opt/homebrew/anaconda3/envs/aaad-poc/lib/python3.9/site-packages/tensordict/tensordict.py:1141, in TensorDictBase()
   1138             future.wait()
   1139         return
-> 1141 def reduce(self, dst, op=dist.ReduceOp.SUM, async_op=False, return_premature=False):
   1142     """Reduces the tensordict across all machines.
   1143 
   1144     Only the process with ``rank`` dst is going to receive the final result.
   1145 
   1146     """
   1147     return self._reduce(dst, op, async_op, return_premature)

AttributeError: module 'torch.distributed' has no attribute 'ReduceOp'

System info

Describe the characteristic of your environment:

• MacOSX 13.2.1
• conda 23.3.1
• python 3.9.18
• torch 2.0.0.post2
• numpy 1.25.2

Checklist

• I have checked that there is no similar issue in the repo (required)
• I have read the documentation (required)
• I have provided a minimal working example to reproduce the bug (required)