Frontend changes for adding per_sample_weights in boundary checker (#…

…1069)

Summary:
Pull Request resolved: #1069

Follow up for D35768276 (7be1fcb): Spin off the frontend changes in a separate Diff to land later to avoid the BC issue.

Reviewed By: RoshanPAN

Differential Revision: D35882262

fbshipit-source-id: 1d58a8aa53a3f9aa1c36d0926e4d9e4274108783

fbgemm_gpu/fbgemm_gpu/split_table_batched_embeddings_ops.py

@@ -599,6 +599,7 @@ def forward(
                 offsets,
                 self.bounds_check_mode_int,
                 self.bounds_check_warning,
+                per_sample_weights,
             )
         self.step += 1
         if len(self.timesteps_prefetched) == 0:
@@ -1972,6 +1973,7 @@ def forward(
                 offsets,
                 self.bounds_check_mode_int,
                 self.bounds_check_warning,
+                per_sample_weights,
             )
         # Note: CPU and CUDA ops use the same interface to facilitate JIT IR
         # generation for CUDA/CPU. For CPU op, we don't need weights_uvm and