diff --git a/include/Types.hpp b/include/Types.hpp
index e895aeae..aa632f66 100644
--- a/include/Types.hpp
+++ b/include/Types.hpp
@@ -51,8 +51,8 @@ enum DataType {
     MLLM_TYPE_COUNT,
 };
 enum ChlType {
-    BHSD = 0,
-    BSHD = 1,
+    BSHD = 0,
+    BHSD = 1,
 };
 
 enum Chl {
diff --git a/src/Tensor.hpp b/src/Tensor.hpp
index 62fd8e00..e2124426 100644
--- a/src/Tensor.hpp
+++ b/src/Tensor.hpp
@@ -158,14 +158,14 @@ class Tensor {
     //    inline int offset(const int n, const int c = 0, const int h = 0,
     //                      const int w = 0) const {
     //        CHECK_GE(n, 0);
-    //        CHECK_LE(n, num());
-    //        CHECK_GE(channels(), 0);
-    //        CHECK_LE(c, channels());
-    //        CHECK_GE(height(), 0);
-    //        CHECK_LE(h, height());
-    //        CHECK_GE(width(), 0);
-    //        CHECK_LE(w, width());
-    //        return ((n * channels() + c) * height() + h) * width() + w;
+    //        CHECK_LE(n, batch());
+    //        CHECK_GE(head(), 0);
+    //        CHECK_LE(c, head());
+    //        CHECK_GE(sequence(), 0);
+    //        CHECK_LE(h, sequence());
+    //        CHECK_GE(dimension(), 0);
+    //        CHECK_LE(w, dimension());
+    //        return ((n * head() + c) * sequence() + h) * dimension() + w;
     //    }
     inline int offset(const int b, const int h = 0, const int s = 0,
                       const int d = 0) const {
@@ -536,6 +536,7 @@ class Tensor {
     // TODO:Name?
 
 
+
     template <class Dtype>
     void fullData(Dtype value) {
         for (int n = 0; n < batch(); ++n) {
diff --git a/src/backends/cpu/CPUKVCache.hpp b/src/backends/cpu/CPUKVCache.hpp
index 3f911d54..380acc4b 100644
--- a/src/backends/cpu/CPUKVCache.hpp
+++ b/src/backends/cpu/CPUKVCache.hpp
@@ -22,7 +22,8 @@ class CPUKVCache final : public Op {
 
 private:
     bool support_multi_thread_ = false;
-    int cache_seq_len_= -INFINITY;
+
+    int cache_seq_len_= -999;
     bool isK_;
 
     int cache_limit_ ;