Check zch config compability for transfer (#3405)

Summary:

Added a check to validate if zch config is compatible between the source model and the target model. This update is to avoid incorrectly MaaS transfer.

For example if the bucket number in the source model is 16 and the number in the target mdoel is 12, even if the zch table size stays the same between these two models, we should apply MPZCH transfer, instead of a noraml transfer, because each row in the source table could map to a different location in the target table.

In this update:
1. We added bucket number into state_dict in the checkpoint
2. We compared the bucket numbers between a source model and a target model  during transfer.
3. If the bucket number in the source cannot be divided by the number in the target, we raise an exception.

Reviewed By: zlzhao1104

Differential Revision: D83580368

Author: chaodengusc

torchrec/distributed/mc_modules.py

@@ -827,6 +827,7 @@ def sharded_parameter_names(self, prefix: str = "") -> Iterator[str]:
                     "_output_segments_tensor",
                     "_current_iter_tensor",
                     "_scalar_logger._scalar_logger_steps",
+                    "_hash_zch_bucket",
                 ]:
                     continue
                 if name in module._non_persistent_buffers_set:

torchrec/modules/hash_mc_modules.py

@@ -305,6 +305,7 @@ def __init__(
 
         self._max_probe = max_probe
         self._buckets = total_num_buckets
+        self.register_buffer("_hash_zch_bucket", torch.tensor(total_num_buckets))
         # Do not need to store in buffer since this is created and consumed
         # at each step https://fburl.com/code/axzimmbx
         self._evicted_indices = []