Remove comm_ops import to workaround torch package issues of lego (#2161)

Summary:
Pull Request resolved: #2161

Pyper release fails because of torch package issues of comm_ops (torch package of transform uses old-old version without set_use_sync_collectives).

PT2Pipeline is not used in prod.

Reviewed By: gnahzg

Differential Revision: D58947470

fbshipit-source-id: 68bea8a1066616f8a53a0ca6a550d9d3008f3028

Author: Ivan Kobzarev

torchrec/distributed/train_pipeline/train_pipelines.py

@@ -28,7 +28,6 @@
 
 import torch
 from torch.autograd.profiler import record_function
-from torchrec.distributed.comm_ops import set_use_sync_collectives
 from torchrec.distributed.dist_data import KJTAllToAllTensorsAwaitable
 from torchrec.distributed.model_parallel import ShardedModule
 from torchrec.distributed.train_pipeline.utils import (
@@ -212,7 +211,8 @@ def progress(self, dataloader_iter: Iterator[In]) -> Out:
         if self._iter == 0:
             # Turn on sync collectives for PT2 pipeline.
             # To have similar logic between compiled/graph_break ranks.
-            set_use_sync_collectives(True)
+            # TODO(ivankobzarev): Call torchrec.distributed.comm_ops.set_use_sync_collectives(True) when torch package issue on import of comm_ops is fixed
+            pass
 
         cc = self._compile_configs