Reland [2] [FBGEMM][MTIA] Support MTIA in DenseTableBatchedEmbeddingBagsCodegen (#2100)

Summary:
Pull Request resolved: #2100

Original diff D58137460, get reverted on D58449527

Will reland in two diffs, one with FBGEMM, one with other changes

Reviewed By: egienvalue

Differential Revision: D58473485

fbshipit-source-id: 53a75989bb05978694963a42c232b378a018dde8

Author: gnahzg

torchrec/distributed/batched_embedding_kernel.py

@@ -939,16 +939,20 @@ def __init__(
         weights_precision = data_type_to_sparse_type(config.data_type)
         fused_params = config.fused_params or {}
         output_dtype = fused_params.get("output_dtype", SparseType.FP32)
+        use_cpu: bool = (
+            device is None
+            or device.type == "cpu"
+            or (not (torch.cuda.is_available() or torch.mtia.is_available()))
+        )
         self._emb_module: DenseTableBatchedEmbeddingBagsCodegen = (
             DenseTableBatchedEmbeddingBagsCodegen(
                 list(zip(self._local_rows, self._local_cols)),
                 feature_table_map=self._feature_table_map,
                 pooling_mode=PoolingMode.NONE,
-                use_cpu=device is None
-                or device.type == "cpu"
-                or not torch.cuda.is_available(),
+                use_cpu=use_cpu,
                 weights_precision=weights_precision,
                 output_dtype=output_dtype,
+                use_mtia=device is not None and device.type == "mtia",
             )
         )
         self._param_per_table: Dict[str, TableBatchedEmbeddingSlice] = dict(
@@ -1376,16 +1380,20 @@ def __init__(
         weights_precision = data_type_to_sparse_type(config.data_type)
         fused_params = config.fused_params or {}
         output_dtype = fused_params.get("output_dtype", SparseType.FP32)
+        use_cpu: bool = (
+            device is None
+            or device.type == "cpu"
+            or (not (torch.cuda.is_available() or torch.mtia.is_available()))
+        )
         self._emb_module: DenseTableBatchedEmbeddingBagsCodegen = (
             DenseTableBatchedEmbeddingBagsCodegen(
                 list(zip(self._local_rows, self._local_cols)),
                 feature_table_map=self._feature_table_map,
                 pooling_mode=self._pooling,
-                use_cpu=device is None
-                or device.type == "cpu"
-                or not torch.cuda.is_available(),
+                use_cpu=use_cpu,
                 weights_precision=weights_precision,
                 output_dtype=output_dtype,
+                use_mtia=device is not None and device.type == "mtia",
             )
         )
         self._param_per_table: Dict[str, TableBatchedEmbeddingSlice] = dict(