fix: Fix ORT backend memory leak (#261)

krishung5 · web-flow · commit 373bd88f7cd8 · 2024-07-15T11:48:31.000-07:00
* Call AllocatorFree

* Handle return
diff --git a/src/onnxruntime.cc b/src/onnxruntime.cc
@@ -872,6 +872,7 @@ ModelState::LoadModel(
          "' on device " + std::to_string(instance_group_device_id) +
          std::string(" with options: ") + std::string(options))
             .c_str());
+    RETURN_IF_ORT_ERROR(ort_api->AllocatorFree(allocator, options));
   }
 #endif  // TRITON_ENABLE_GPU
 

Original file line number	Diff line number	Diff line change
`@@ -872,6 +872,7 @@ ModelState::LoadModel(`
`872`	`872`	`"' on device " + std::to_string(instance_group_device_id) +`
`873`	`873`	`std::string(" with options: ") + std::string(options))`
`874`	`874`	`.c_str());`
	`875`	`+ RETURN_IF_ORT_ERROR(ort_api->AllocatorFree(allocator, options));`
`875`	`876`	`}`
`876`	`877`	`#endif // TRITON_ENABLE_GPU`
`877`	`878`