Update sparse_ops.py to use generic gpu target fbgemm_gpu:input_combine to support both nvidia and AMD (#2905)

Summary:
Pull Request resolved: #2905

Update sparse_ops.py to use generic gpu target fbgemm_gpu:input_combine to support both nvidia and AMD

Reviewed By: Yuzhen11

Differential Revision: D60271751

fbshipit-source-id: 17aeb321ebf805589c6e8f3f3698220725521a2e

Author: joebos

fbgemm_gpu/fbgemm_gpu/sparse_ops.py

@@ -26,7 +26,6 @@
         torch.ops.load_library(
             "//deeplearning/fbgemm/fbgemm_gpu/codegen:embedding_ops_hip"
         )
-        torch.ops.load_library("//deeplearning/fbgemm/fbgemm_gpu:input_combine_hip")
         torch.ops.load_library(
             "//deeplearning/fbgemm/fbgemm_gpu/codegen:index_select_ops_hip"
         )
@@ -36,7 +35,8 @@
             "//deeplearning/fbgemm/fbgemm_gpu:merge_pooled_embeddings"
         )
         torch.ops.load_library("//deeplearning/fbgemm/fbgemm_gpu/codegen:embedding_ops")
-        torch.ops.load_library("//deeplearning/fbgemm/fbgemm_gpu:input_combine")
+
+    torch.ops.load_library("//deeplearning/fbgemm/fbgemm_gpu:input_combine")
 
     torch.ops.load_library("//deeplearning/fbgemm/fbgemm_gpu:sparse_ops_cpu")
     torch.ops.load_library(