From 110ac6a5e13f828b6086099888fc6cef8de095d5 Mon Sep 17 00:00:00 2001
From: xiaying <xiaotang.jxt@alibaba-inc.com>
Date: Fri, 3 Jan 2025 18:42:00 +0800
Subject: [PATCH] LLM:Bugfix: Fix compile bug for mllm

---
 docs/transformers/llm.md            | 2 +-
 transformers/llm/engine/src/llm.cpp | 6 ++----
 2 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/docs/transformers/llm.md b/docs/transformers/llm.md
index bdd40739f..109a6b43f 100644
--- a/docs/transformers/llm.md
+++ b/docs/transformers/llm.md
@@ -107,7 +107,7 @@ options:
 ```
 - 需要开启音频功能时，增加相关编译宏
 ```
--DLLM_SUPPORT_AUDIO=true
+-DLLM_SUPPORT_AUDIO=true -DMNN_BUILD_AUDIO=true
 ```
 
 #### mac / linux / windows
diff --git a/transformers/llm/engine/src/llm.cpp b/transformers/llm/engine/src/llm.cpp
index 293e1af1b..264304650 100644
--- a/transformers/llm/engine/src/llm.cpp
+++ b/transformers/llm/engine/src/llm.cpp
@@ -1085,7 +1085,7 @@ std::vector<int> Mllm::vision_process(const std::string& file) {
         image_embedding = mul_module_->forward(image);
     }
     auto et    = std::chrono::system_clock::now();
-    vision_us_ = std::chrono::duration_cast<std::chrono::microseconds>(et - st).count();
+    mState.vision_us_ = std::chrono::duration_cast<std::chrono::microseconds>(et - st).count();
     mul_embeddings_.push_back(image_embedding);
     int visual_len = image_embedding->getInfo()->dim[0];
     std::vector<int> img_ids(visual_len, img_pad_);
@@ -1115,7 +1115,7 @@ std::vector<int> Mllm::audio_process(const std::string& file) {
     auto audio_embedding = mul_module_->forward(input_features);
     audio_embedding = _Permute(audio_embedding, {1, 0, 2});
     auto et         = std::chrono::system_clock::now();
-    audio_us_       = std::chrono::duration_cast<std::chrono::microseconds>(et - st).count();
+    mState.audio_us_       = std::chrono::duration_cast<std::chrono::microseconds>(et - st).count();
     mul_embeddings_.push_back(audio_embedding);
     int embed_len = audio_embedding->getInfo()->dim[0];
     std::vector<int> audio_ids(embed_len, audio_pad_);
@@ -1172,11 +1172,9 @@ std::vector<int> Mllm::multimode_process(const std::string& mode, std::string in
                 file.close();
             } else {
                 std::cerr << "Unable to open file to write." << std::endl;
-                exit(0);
             }
         } else {
             std::cerr << "Failed to download file. Status code: " << (res ? res->status : 0) << std::endl;
-            exit(0);
         }
     }
     if (mode == "img" && config_->is_visual()) {