remove d2d copies (#64)

* remove d2d copies
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

* cleanup
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

---------
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

transformer_engine/pytorch/csrc/ts_fp8_op.cpp

@@ -30,7 +30,7 @@ at::Tensor cast_to_fp8_ts(const at::Tensor &input,
                                    amax[0][fp8_tensor],
                                    scale_inv[fp8_tensor],
                                    otype_arg);
-  return output.clone();
+  return output;
 }
 
 at::Tensor cast_from_fp8_ts(const at::Tensor &input,
@@ -44,7 +44,7 @@ at::Tensor cast_from_fp8_ts(const at::Tensor &input,
                                     scale_inv[fp8_tensor],
                                     itype_arg,
                                     otype_arg);
-  return output.clone();
+  return output;
 }
 
 at::Tensor fp8_gelu_ts(at::Tensor input,
@@ -59,7 +59,7 @@ at::Tensor fp8_gelu_ts(at::Tensor input,
                                amax[0][fp8_tensor],
                                scale_inv[fp8_tensor],
                                otype_arg);
-  return output.clone();
+  return output;
 }
 
 at::Tensor te_gemm_ts(at::Tensor A,
@@ -92,20 +92,18 @@ at::Tensor te_gemm_ts(at::Tensor A,
   bool accumulate_arg = static_cast<bool>(accumulate);
   bool use_split_accumulator_arg = static_cast<bool>(use_split_accumulator);
 
-  at::Tensor A_scale_inverse_arg = A_scale_inverse.clone();
   if (A_scale_inverse.numel())
-    A_scale_inverse_arg = A_scale_inverse[A_fp8_tensor];
+    A_scale_inverse = A_scale_inverse[A_fp8_tensor];
 
-  at::Tensor B_scale_inverse_arg = B_scale_inverse.clone();
   if (B_scale_inverse.numel())
-    B_scale_inverse_arg = B_scale_inverse[B_fp8_tensor];
+    B_scale_inverse = B_scale_inverse[B_fp8_tensor];
 
   te_gemm(A,
-          A_scale_inverse_arg,
+          A_scale_inverse,
           A_type_arg,
           transa_arg,
           B,
-          B_scale_inverse_arg,
+          B_scale_inverse,
           B_type_arg,
           transb_arg,
           D,
@@ -141,7 +139,7 @@ at::Tensor layernorm_fwd_fp8_inf_ts(const at::Tensor &input,
                                             scale_inv,
                                             otype_arg);
 
-  return output.clone();
+  return output;
 }
 
 at::Tensor layernorm_fwd_inf_ts(const at::Tensor &input,
@@ -155,7 +153,7 @@ at::Tensor layernorm_fwd_inf_ts(const at::Tensor &input,
                                         bias,
                                         eps_float);
 
-  return output.clone();
+  return output;
 }
 
 TORCH_LIBRARY(tex_ts, m) {