Remove the hard-coded value for bucket buffer (#3581)

Summary:
Pull Request resolved: #3581

As title.

Differential Revision: D87966872

fbshipit-source-id: fb8e26b3aa59d57b3015be9ff788882dc628576a

Author: chaodengusc

torchrec/modules/hash_mc_modules.py

@@ -185,6 +185,7 @@ class HashZchManagedCollisionModule(ManagedCollisionModule):
 
     IDENTITY_BUFFER: str = "_hash_zch_identities"
     METADATA_BUFFER: str = "_hash_zch_metadata"
+    BUCKET_BUFFER: str = "_hash_zch_bucket"
 
     table_name_on_device_remapped_ids_dict: Dict[
         str, torch.Tensor
@@ -308,7 +309,10 @@ def __init__(
 
         self._max_probe = max_probe
         self._buckets = total_num_buckets
-        self.register_buffer("_hash_zch_bucket", torch.tensor([[total_num_buckets]]))
+        self.register_buffer(
+            HashZchManagedCollisionModule.BUCKET_BUFFER,
+            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 = []