diff --git a/.pins/et-pin.txt b/.pins/et-pin.txt
index 4411b049c..8e52c115d 100644
--- a/.pins/et-pin.txt
+++ b/.pins/et-pin.txt
@@ -1 +1 @@
-ff4e9edc55c86953ae33dbbaaacab8c9949dcb74
+172574a6be5910a4609e4ed1bef2b6b8475ddb3d
diff --git a/runner/run.cpp b/runner/run.cpp
index 16521ad26..5d20bcc62 100644
--- a/runner/run.cpp
+++ b/runner/run.cpp
@@ -203,9 +203,9 @@ float* forward(Transformer* transformer, int token, int pos) {
                              .to(torch::kCPU);
   auto logits = result[0].data_ptr();
 #else // __ET_MODEL__
-  ManagedTensor pos_managed(pos_buffer, sizeof(int64_t), {1}, ScalarType::Long);
+  ManagedTensor pos_managed(pos_buffer, {1}, ScalarType::Long);
   ManagedTensor tokens_managed(
-      token_buffer, sizeof(int64_t), {1, 1}, ScalarType::Long);
+      token_buffer, {1, 1}, ScalarType::Long);
   std::vector<EValue> inputs;
   auto tmp1 = EValue(tokens_managed.get_aliasing_tensor());
   auto tmp2 = EValue(pos_managed.get_aliasing_tensor());