Add max_enqueued_batches option for model servers.

PiperOrigin-RevId: 639219039
tensorflow · Jun 1, 2024 · 67a2dcb · 67a2dcb
1 parent d914192
commit 67a2dcb
Show file tree

Hide file tree

Showing 2 changed files with 0 additions and 5 deletions.
diff --git a/tensorflow_serving/servables/tensorflow/tfrt_saved_model_factory.cc b/tensorflow_serving/servables/tensorflow/tfrt_saved_model_factory.cc
@@ -234,8 +234,6 @@ absl::Status TfrtSavedModelFactory::CreateTfrtSavedModelWithMetadata(
   compile_options.use_gpu_compile_and_execute_op =
       config_.tfrt_use_fused_gpu_op();
   compile_options.min_num_batch_threads = config_.tfrt_min_num_batch_threads();
-  compile_options.min_max_enqueued_batches =
-      config_.tfrt_min_max_enqueued_batches();
 
   options.graph_execution_options.run_placer_grappler_on_functions =
       config_.run_placer_grappler_on_functions();

diff --git a/tensorflow_serving/servables/tensorflow/tfrt_saved_model_source_adapter.proto b/tensorflow_serving/servables/tensorflow/tfrt_saved_model_source_adapter.proto
@@ -208,9 +208,6 @@ message TfrtSavedModelConfig {
   // Whether to enable paging. This should only be true when using Pathways
   // backend.
   bool enable_paging = 2022;
-
-  // The minimum of the maximum number of outstanding enqueue batches
-  int64 tfrt_min_max_enqueued_batches = 2023;
 }
 
 // Config proto for TfrtSavedModelSourceAdapter.