Merge branch 'master' into ci/docs/add-mkdocs

Author: aktech

RELEASE.md

@@ -1,5 +1,19 @@
 <!-- disableFinding(HEADING_REPEAT_H1) -->
 
+# Release 2.19.1
+
+## Major Features and Improvements
+
+* No major features or improvements.
+
+## Breaking Changes
+
+* No breaking changes.
+
+## Bug Fixes and Other Changes
+
+* This release is based on TF version 2.19.1.
+
 # Release 2.19.0
 
 ## Major Features and Improvements

WORKSPACE

@@ -51,6 +51,13 @@ http_archive(
     ],
 )
 
+http_archive(
+    name = "rules_cc",
+    sha256 = "b8b918a85f9144c01f6cfe0f45e4f2838c7413961a8ff23bc0c6cdf8bb07a3b6",
+    strip_prefix = "rules_cc-0.1.5",
+    url = "https://github.com/bazelbuild/rules_cc/releases/download/0.1.5/rules_cc-0.1.5.tar.gz",
+)
+
 http_archive(
     name = "rules_python",
     sha256 = "84aec9e21cc56fbc7f1335035a71c850d1b9b5cc6ff497306f84cced9a769841",

tensorflow_serving/apis/BUILD

@@ -4,6 +4,8 @@
 
 # Placeholder: load java_proto_library
 
+load("@rules_cc//cc:cc_library.bzl", "cc_library")
+
 # Placeholder: load py_library
 # Placeholder: load kt_jvm_proto_library
 load("@org_tensorflow//tensorflow/core/platform:build_config.bzl", "tf_jspb_proto_library", "tf_pyclif_proto_library")

tensorflow_serving/apis/predict.proto

@@ -71,6 +71,10 @@ message PredictRequest {
     // response if the model stops at them. The model may stop at other tokens,
     // but will not return them in the response.
     repeated int64 return_stoptokens = 4;
+
+    // When set and supported by servable, the model server will cache the
+    // prefix of request up to this length.
+    optional int64 max_cache_length = 6;
   }
 
   optional RequestOptions request_options = 7;

tensorflow_serving/batching/BUILD

@@ -1,5 +1,8 @@
 # Description: Tensorflow Serving batching.
 
+load("@rules_cc//cc:cc_library.bzl", "cc_library")
+load("@rules_cc//cc:cc_test.bzl", "cc_test")
+
 package(
     default_visibility = ["//tensorflow_serving:internal"],
     features = ["-layering_check"],
@@ -25,6 +28,7 @@ cc_library(
     visibility = ["//visibility:private"],
     deps = [
         ":batch_scheduler_retrier",
+        "@com_google_absl//absl/synchronization",
         "@com_google_absl//absl/types:optional",
         "@org_tensorflow//tensorflow/core:lib",
         "@org_tensorflow//tensorflow/core/kernels/batching_util:batch_scheduler",
@@ -39,6 +43,7 @@ cc_test(
     deps = [
         ":streaming_batch_scheduler",
         "//tensorflow_serving/core/test_util:test_main",
+        "@com_google_absl//absl/synchronization",
         "@org_tensorflow//tensorflow/core:lib",
         "@org_tensorflow//tensorflow/core:protos_all_cc",
         "@org_tensorflow//tensorflow/core:test",
@@ -77,6 +82,7 @@ cc_library(
         "//tensorflow_serving/servables/tensorflow:serving_session",
         "//tensorflow_serving/util:hash",
         "@com_google_absl//absl/container:fixed_array",
+        "@com_google_absl//absl/synchronization",
         "@com_google_absl//absl/types:optional",
         "@org_tensorflow//tensorflow/core:core_cpu",
         "@org_tensorflow//tensorflow/core:framework",
@@ -148,6 +154,7 @@ cc_test(
     deps = [
         ":batch_scheduler_retrier",
         "//tensorflow_serving/core/test_util:test_main",
+        "@com_google_absl//absl/synchronization",
         "@org_tensorflow//tensorflow/core:lib",
         "@org_tensorflow//tensorflow/core:protos_all_cc",
         "@org_tensorflow//tensorflow/core:test",
@@ -217,6 +224,7 @@ cc_library(
         ":incremental_barrier",
         "@com_google_absl//absl/container:flat_hash_map",
         "@com_google_absl//absl/container:inlined_vector",
+        "@com_google_absl//absl/synchronization",
         "@org_tensorflow//tensorflow/core:framework",
         "@org_tensorflow//tensorflow/core:lib",
         "@org_tensorflow//tensorflow/core/kernels/batching_util:batch_scheduler",

tensorflow_serving/batching/batch_scheduler_retrier_test.cc

@@ -20,9 +20,9 @@ limitations under the License.
 #include <utility>
 
 #include <gtest/gtest.h>
+#include "absl/synchronization/notification.h"
 #include "tensorflow/core/kernels/batching_util/fake_clock_env.h"
 #include "tensorflow/core/lib/core/errors.h"
-#include "tensorflow/core/lib/core/notification.h"
 #include "tensorflow/core/lib/core/status.h"
 #include "tensorflow/core/lib/core/status_test_util.h"
 #include "tensorflow/core/platform/macros.h"
@@ -149,7 +149,7 @@ TEST(BatchSchedulerRetrierTest, MaxTime) {
           options, std::move(stubborn_scheduler), &retrier));
 
       const bool expect_success = max_attempts >= num_attempts_to_succeed;
-      Notification done;
+      absl::Notification done;
       std::unique_ptr<Thread> run_retrier(Env::Default()->StartThread(
           {}, "RunRetrier",
           [&retrier, &expect_success, &done]() {
@@ -191,7 +191,7 @@ TEST(BatchSchedulerRetrierTest, RetryDelay) {
   TF_CHECK_OK(BatchSchedulerRetrier<FakeTask>::Create(
       options, std::move(stubborn_scheduler), &retrier));
 
-  Notification done;
+  absl::Notification done;
   std::unique_ptr<Thread> run_retrier(Env::Default()->StartThread(
       {}, "RunRetrier",
       [&retrier, &done]() {

tensorflow_serving/batching/batching_session.cc

@@ -26,13 +26,13 @@ limitations under the License.
 #include <vector>
 
 #include "absl/container/fixed_array.h"
+#include "absl/synchronization/notification.h"
 #include "tensorflow/core/framework/cost_graph.pb.h"
 #include "tensorflow/core/framework/tensor.h"
 #include "tensorflow/core/framework/tensor_shape.h"
 #include "tensorflow/core/framework/tensor_util.h"
 #include "tensorflow/core/kernels/batching_util/input_split_metadata.h"
 #include "tensorflow/core/lib/core/errors.h"
-#include "tensorflow/core/lib/core/notification.h"
 #include "tensorflow/core/lib/core/status.h"
 #include "tensorflow/core/lib/gtl/cleanup.h"
 #include "tensorflow/core/lib/monitoring/counter.h"
@@ -435,7 +435,7 @@ absl::Status BatchingSession::InternalRun(
 
   outputs->clear();
 
-  Notification done;
+  absl::Notification done;
   absl::Status status;
   auto task = std::unique_ptr<BatchingSessionTask>(new BatchingSessionTask);
   task->enqueue_time_micros = EnvTime::NowMicros();

tensorflow_serving/batching/batching_session.h

@@ -26,6 +26,7 @@ limitations under the License.
 #include <utility>
 #include <vector>
 
+#include "absl/synchronization/notification.h"
 #include "absl/types/optional.h"
 #include "tensorflow/core/kernels/batching_util/basic_batch_scheduler.h"
 #include "tensorflow/core/kernels/batching_util/batch_scheduler.h"
@@ -180,7 +181,7 @@ struct BatchingSessionTask : public BatchTask {
 
   // Fields populated when a task is processed (as part of a batch), and
   // returned by BatchingSession when a task is complete.
-  Notification* done;
+  absl::Notification* done;
   Status* status;
   std::vector<Tensor>* outputs;
   RunMetadata* run_metadata;

tensorflow_serving/batching/batching_session_test.cc

@@ -375,7 +375,7 @@ TEST_P(BatchingSessionTest, BatchingWithLargeBatch) {
     Tensor expected_output1 =
         test::AsTensor<float>({4.5, 5, 5.5, 6}, {1, 2, 2});
     std::vector<Tensor> output1;
-    Notification notify;
+    absl::Notification notify;
     std::unique_ptr<Thread> first_request_thread(
         Env::Default()->StartThread(ThreadOptions(), "first_request", [&] {
           auto status =
@@ -991,7 +991,7 @@ TEST_P(BatchingSessionTest, EnqueuedLongerThanTimeout) {
   ASSERT_FALSE(scheduler == nullptr);
 
   // Enqueue a request with a timeout specified via RunOptions.
-  Notification request_returned;
+  absl::Notification request_returned;
   auto issue_request = [&batching_session, &request_returned] {
     Tensor input = test::AsTensor<float>({100.0f, 42.0f}, {2});
     RunOptions run_options;

tensorflow_serving/batching/streaming_batch_scheduler.h

@@ -17,15 +17,16 @@ limitations under the License.
 #define TENSORFLOW_SERVING_BATCHING_STREAMING_BATCH_SCHEDULER_H_
 
 #include <stddef.h>
+
 #include <algorithm>
 #include <functional>
 #include <memory>
 #include <string>
 #include <utility>
 
+#include "absl/synchronization/notification.h"
 #include "tensorflow/core/kernels/batching_util/batch_scheduler.h"
 #include "tensorflow/core/lib/core/errors.h"
-#include "tensorflow/core/lib/core/notification.h"
 #include "tensorflow/core/lib/core/status.h"
 #include "tensorflow/core/lib/core/threadpool.h"
 #include "tensorflow/core/platform/cpu_info.h"
@@ -283,7 +284,7 @@ class SingleTaskScheduler {
   uint64_t last_task_time_ = 0;
 
   // A notification for stopping the thread, during destruction.
-  Notification stop_;
+  absl::Notification stop_;
 
   // The name of 'thread_'.
   const string thread_name_;