resolve conflicts and fix UT bugs.

Author: minghaoBD

paddle/fluid/operators/fused/fused_layernorm_residual_dropout_bias.h

@@ -678,15 +678,15 @@ __global__ __launch_bounds__(THREADS_PER_CTA) void fused_fast_ln_fwd_kernel(
         x[it][jt] = static_cast<T>(static_cast<U>(gamma[it][jt]) * tmp +
                                    static_cast<U>(beta[it][jt]));
 
-        if (std::is_same<T, int8_t>::value)
+        if (std::is_same<OutType, int8_t>::value)
           x_output[it][jt] = __float2int_rn(static_cast<float>(x[it][jt]) *
                                             quant_in_scale_data);
       }
     }
 
 #pragma unroll
     for (int it = 0, col = c; it < LDGS; it++) {
-      if (std::is_same<T, int8_t>::value) {
+      if (std::is_same<OutType, int8_t>::value) {
         phi::Store<OutType, VecSize>(
             x_output[it], y_ptr + row * ELTS_PER_ROW + col * VecSize);
       } else {

paddle/fluid/operators/fused/fused_multi_transformer_int8_op.cu

@@ -1,5 +1,4 @@
 /* Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
- * Copyright (c) 2011-2021, NVIDIA CORPORATION.  All rights reserved.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -12,9 +11,6 @@ distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
-// This file has been adapted from FasterTransformer file:
-// https://github.com/NVIDIA/FasterTransformer/blob/v4.0/fastertransformer/cuda/masked_multihead_attention.cu
-// We add License in the head.
 
 #include "paddle/fluid/operators/fused/attn_gemm_int8.h"
 #include "paddle/fluid/operators/fused/fused_multi_transformer_op.h"