Remove explicit dereferencing for TesnorPtr converted implicitly to E…

…Value. (pytorch#5278) Summary: Pull Request resolved: pytorch#5278 Reviewed By: kirklandsign Differential Revision: D62512518 fbshipit-source-id: c1f32dd398cb58833ca3fa95b0cd1ab5c9984de9
dpalmasan · Sep 11, 2024 · 338ef26 · 338ef26
1 parent d689722
commit 338ef26
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/examples/qualcomm/oss_scripts/llama2/runner/runner.cpp b/examples/qualcomm/oss_scripts/llama2/runner/runner.cpp
@@ -145,7 +145,7 @@ Result<torch::executor::Tensor> Runner::run_model_step(
   token->mutable_data_ptr<int32_t>()[0] = input_token;
 
   // inputs:[tokens, start_pos, atten_mask, k_cache, v_cache]
-  auto outputs_res = module_->forward({*token, *start_pos, *atten_mask});
+  auto outputs_res = module_->forward({token, start_pos, atten_mask});
   ET_CHECK_OK_OR_RETURN_ERROR(outputs_res.error());
 
   // TODO: need to handle batch size != 1

diff --git a/extension/llm/runner/text_decoder_runner.cpp b/extension/llm/runner/text_decoder_runner.cpp
@@ -42,7 +42,7 @@ ::executorch::runtime::Result<exec_aten::Tensor> TextDecoderRunner::step(
     TensorPtr& start_pos) {
   // ET_LOG(Info, "Input token %" PRIu64, input_token);
   if (use_kv_cache_) {
-    auto outputs_res = module_->forward({*tokens, *start_pos});
+    auto outputs_res = module_->forward({tokens, start_pos});
     ET_CHECK_OK_OR_RETURN_ERROR(outputs_res.error());
     ET_CHECK_MSG(
         outputs_res.get().size() == 1,