From 612d5da08cf4375d7f22fcdc298a0524ce74b03b Mon Sep 17 00:00:00 2001
From: zhoutianzi666 <39978853+zhoutianzi666@users.noreply.github.com>
Date: Fri, 24 Feb 2023 10:53:16 +0800
Subject: [PATCH] [Paddle-TRT] Fix QkvToContextPluginDynamic bug (#50715)

* fix multihead

* fix multihead
---
 paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.cu | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.cu b/paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.cu
index 731441463df..fb2b60a2de5 100644
--- a/paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.cu
+++ b/paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.cu
@@ -479,6 +479,8 @@ int QkvToContextPluginDynamic::enqueue(
     const half *input1_data = static_cast<const half *>(qk_bias);
     // BxSx3xNxH => tptr: 3xBxNxSxH.
     if (need_padding) {
+      PADDLE_ENFORCE_GPU_SUCCESS(
+          cudaMemsetAsync(tptr, 0, sizeof(half) * input_num, stream));
       TransposePadding(input0_data,
                        tptr,
                        batch,
-- 
GitLab