remove invalid warning log

frontend/server/src/main/java/org/pytorch/serve/ModelServer.java

@@ -38,6 +38,7 @@
 import org.pytorch.serve.archive.model.ModelNotFoundException;
 import org.pytorch.serve.grpcimpl.GRPCInterceptor;
 import org.pytorch.serve.grpcimpl.GRPCServiceFactory;
+import org.pytorch.serve.http.messages.RegisterModelRequest;
 import org.pytorch.serve.metrics.MetricCache;
 import org.pytorch.serve.metrics.MetricManager;
 import org.pytorch.serve.servingsdk.ModelServerEndpoint;
@@ -214,11 +215,6 @@ private void initModelStore() throws InvalidSnapshotException, IOException {
                             if (marMinWorkers > 0 && marMaxWorkers >= marMinWorkers) {
                                 minWorkers = marMinWorkers;
                                 maxWorkers = marMaxWorkers;
-                            } else {
-                                logger.warn(
-                                        "Invalid model config in mar, minWorkers:{}, maxWorkers:{}",
-                                        marMinWorkers,
-                                        marMaxWorkers);
                             }
                         }
                         modelManager.updateModel(
@@ -266,8 +262,8 @@ private void initModelStore() throws InvalidSnapshotException, IOException {
                                 modelName,
                                 null,
                                 null,
-                                1,
-                                100,
+                                -1 * RegisterModelRequest.DEFAULT_BATCH_SIZE,
+                                -1 * RegisterModelRequest.DEFAULT_MAX_BATCH_DELAY,
                                 configManager.getDefaultResponseTimeout(),
                                 defaultModelName,
                                 false,

frontend/server/src/main/java/org/pytorch/serve/http/messages/RegisterModelRequest.java

@@ -55,8 +55,7 @@ public RegisterModelRequest(QueryStringDecoder decoder) {
                         "initial_workers",
                         ConfigManager.getInstance().getConfiguredDefaultWorkersPerModel());
         synchronous = Boolean.parseBoolean(NettyUtils.getParameter(decoder, "synchronous", "true"));
-        responseTimeout =
-                NettyUtils.getIntParameter(decoder, "response_timeout", -1 * DEFAULT_BATCH_SIZE);
+        responseTimeout = NettyUtils.getIntParameter(decoder, "response_timeout", -1);
         modelUrl = NettyUtils.getParameter(decoder, "url", null);
         s3SseKms = Boolean.parseBoolean(NettyUtils.getParameter(decoder, "s3_sse_kms", "false"));
     }

frontend/server/src/main/java/org/pytorch/serve/wlm/ModelManager.java

@@ -76,8 +76,8 @@ public ModelArchive registerModel(String url, String defaultModelName)
                 null,
                 null,
                 null,
-                1,
-                100,
+                -1 * RegisterModelRequest.DEFAULT_BATCH_SIZE,
+                -1 * RegisterModelRequest.DEFAULT_MAX_BATCH_DELAY,
                 configManager.getDefaultResponseTimeout(),
                 defaultModelName,
                 false,

frontend/server/src/main/java/org/pytorch/serve/wlm/WorkerThread.java

@@ -293,7 +293,9 @@ public void run() {
                 logger.debug("Shutting down the thread .. Scaling down.");
             } else {
                 logger.debug(
-                        "Backend worker monitoring thread interrupted or backend worker process died.",
+                        "Backend worker monitoring thread interrupted or backend worker process died., responseTimeout:"
+                                + responseTimeout
+                                + "sec",
                         e);
             }
         } catch (WorkerInitializationException e) {
@@ -586,7 +588,8 @@ public void channelRead0(ChannelHandlerContext ctx, ModelWorkerResponse msg) {
             try {
                 replies.offer(msg, responseTimeout, TimeUnit.SECONDS);
             } catch (InterruptedException | NullPointerException e) {
-                logger.error("Failed to offer reply", e);
+                logger.error(
+                        "Failed to offer reply, responseTimeout:" + responseTimeout + "sec", e);
                 throw new IllegalStateException("Reply queue is full.");
             }
         }

test/pytest/test_model_config.py

@@ -96,48 +96,42 @@ def create_mar_file(work_dir, model_archiver, model_name):
     mar_file_path.unlink(missing_ok=True)
 
 
-def register_model(mar_file_path, model_store, params, torchserve):
+@pytest.fixture(scope="module", name="base_model_name")
+def register_model(mar_file_path, model_store, torchserve):
     shutil.copy(mar_file_path, model_store)
 
     file_name = Path(mar_file_path).name
 
-    model_name = Path(file_name).stem
+    base_model_name = Path(file_name).stem
 
-    params = params + (
-        ("model_name", model_name),
+    params = (
+        ("model_name", f"{base_model_name}_b4"),
         ("url", file_name),
+        ("initial_workers", "2"),
+        ("synchronous", "true"),
     )
 
     test_utils.reg_resp = test_utils.register_model_with_params(params)
-    return model_name
-
 
-@pytest.mark.skip(reason="Flaky on Regression GPU")
-def test_register_model_with_batch_size(mar_file_path, model_store, torchserve):
     params = (
+        ("model_name", f"{base_model_name}_b2"),
+        ("url", file_name),
         ("initial_workers", "2"),
         ("synchronous", "true"),
         ("batch_size", "2"),
     )
 
-    model_name = register_model(mar_file_path, model_store, params, torchserve)
-
-    describe_resp = test_utils.describe_model(model_name, "1.0")
+    test_utils.reg_resp = test_utils.register_model_with_params(params)
+    yield base_model_name
 
-    assert describe_resp[0]["batchSize"] == 2
 
-    test_utils.unregister_model(model_name)
+def test_register_model_with_batch_size(base_model_name):
+    describe_resp = test_utils.describe_model(f"{base_model_name}_b2", "1.0")
 
+    assert describe_resp[0]["batchSize"] == 2
 
-def test_register_model_without_batch_size(mar_file_path, model_store, torchserve):
-    params = (
-        ("initial_workers", "2"),
-        ("synchronous", "true"),
-    )
-    model_name = register_model(mar_file_path, model_store, params, torchserve)
 
-    describe_resp = test_utils.describe_model(model_name, "1.0")
+def test_register_model_without_batch_size(base_model_name):
+    describe_resp = test_utils.describe_model(f"{base_model_name}_b4", "1.0")
 
     assert describe_resp[0]["batchSize"] == 4
-
-    test_utils.unregister_model(model_name)