Fix decoupled batch statistics to account for implicit batch size

triton-inference-server · May 22, 2024 · 89512e6 · 89512e6
1 parent bc16e1a
commit 89512e6
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/python_be.cc b/src/python_be.cc
@@ -1236,7 +1236,7 @@ ModelInstanceState::ProcessRequests(
   uint64_t compute_end_ns = 0;
   SET_TIMESTAMP(compute_end_ns);
   reporter.SetComputeEndNs(compute_end_ns);
-  reporter.SetBatchStatistics(request_count);
+  reporter.SetBatchStatistics(total_batch_size);
 
   if (response_batch.data_->has_error) {
     if (response_batch.data_->is_error_set) {