diff --git a/.github/workflows/build-test.yml b/.github/workflows/build-test.yml
index 0bd8fa81..f1ff7853 100644
--- a/.github/workflows/build-test.yml
+++ b/.github/workflows/build-test.yml
@@ -70,10 +70,16 @@ jobs:
             unzip $PACAGE_FILE
             cd $PACAGE_DIR
             ./script/model_test.sh ${{ matrix.model }}
-      - name: windows-test
+      - name: windows-download
         if: matrix.os == 'windows-latest'
         run: |
             cd workspace
             7z x windows-package.zip
             cd windows-package
-            ./script/model_test.ps1 ${{ matrix.model }}
\ No newline at end of file
+            echo ${{ matrix.model }}
+            ./script/model_download.ps1 ${{ matrix.model }}
+      - name: windows-test
+        if: matrix.os == 'windows-latest'
+        run: |
+            cd build
+            .\Release\cli_demo -m ..\${{ matrix.model }}
\ No newline at end of file
diff --git a/include/llm.hpp b/include/llm.hpp
index 9f0eba8d..29ab9745 100644
--- a/include/llm.hpp
+++ b/include/llm.hpp
@@ -50,6 +50,7 @@ class Llm {
     // gen info
     int gen_seq_len_ = 0;
     int all_seq_len_ = 0;
+    int max_seq_len_ = 256;
     float load_progress_ = 0.f;
 private:
     // MNN Modules
diff --git a/src/llm.cpp b/src/llm.cpp
index a3a56d6f..a133f903 100644
--- a/src/llm.cpp
+++ b/src/llm.cpp
@@ -109,7 +109,7 @@ std::string Llm::response(const std::string& query, std::ostream* os) {
     int token = forward(input_ids);
     std::string output_str = decode(token);
     *os << output_str << std::flush;
-    while (true) {
+    while (gen_seq_len_ < max_seq_len_) {
         token = forward({token});
         if (is_stop(token)) {
             *os << std::endl << std::flush;