Fix is_using_virtual_table bug in kernel

Differential Revision: D86233647

Author: EddyLXJ

torchrec/distributed/batched_embedding_kernel.py

@@ -194,7 +194,7 @@ def _populate_ssd_tbe_params(config: GroupedEmbeddingConfig) -> Dict[str, Any]:
         )
 
     # populate init min and max
-    if config.is_using_virtual_table:
+    if config.is_using_virtual_table():
         _generate_init_range_for_virtual_tables(ssd_tbe_params, config)
 
     if (
@@ -242,7 +242,7 @@ def _populate_ssd_tbe_params(config: GroupedEmbeddingConfig) -> Dict[str, Any]:
             )
             ssd_tbe_params["cache_sets"] = int(max_cache_sets)
 
-    if "kvzch_eviction_tbe_config" in fused_params and config.is_using_virtual_table:
+    if "kvzch_eviction_tbe_config" in fused_params and config.is_using_virtual_table():
         ssd_tbe_params["kvzch_eviction_tbe_config"] = fused_params.get(
             "kvzch_eviction_tbe_config"
         )
@@ -1975,7 +1975,7 @@ def __init__(
             len({table.embedding_dim for table in config.embedding_tables}) == 1
         ), "Currently we expect all tables in SSD TBE to have the same embedding dimension."
         assert (
-            config.is_using_virtual_table
+            config.is_using_virtual_table()
         ), "Try to create ZeroCollisionKeyValueEmbedding for non virtual tables"
         assert embedding_cache_mode == config.enable_embedding_update, (
             f"Embedding_cache kernel is {embedding_cache_mode} "
@@ -2889,7 +2889,7 @@ def __init__(
             len({table.embedding_dim for table in config.embedding_tables}) == 1
         ), "Currently we expect all tables in SSD TBE to have the same embedding dimension."
         assert (
-            config.is_using_virtual_table
+            config.is_using_virtual_table()
         ), "Try to create ZeroCollisionKeyValueEmbeddingBag for non virtual tables"
 
         for table in config.embedding_tables: