diff --git a/cpp/tensorrt_llm/batch_manager/promptTuningBuffers.cpp b/cpp/tensorrt_llm/batch_manager/promptTuningBuffers.cpp
index e9ae8cec93e..1cf73a2c0d2 100644
--- a/cpp/tensorrt_llm/batch_manager/promptTuningBuffers.cpp
+++ b/cpp/tensorrt_llm/batch_manager/promptTuningBuffers.cpp
@@ -117,7 +117,6 @@ void PromptTuningBuffers::fill(RequestVector const& contextRequests, RequestVect
     runtime::BufferManager const& manager, bool packed)
 {
     NVTX3_SCOPED_RANGE_WITH_NAME(range, "PromptTuningBuffers::fill");
-    manager.setZero(*mPromptTuningParams.embeddingTable);
 
     auto const numContextRequests = static_cast<SizeType32>(contextRequests.size());