adding source of truth

pytorch · Sep 23, 2024 · 970db0e · 970db0e
1 parent 9a00295
commit 970db0e
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 21 deletions.
diff --git a/frontend/archive/src/main/java/org/pytorch/serve/archive/model/ModelConfig.java b/frontend/archive/src/main/java/org/pytorch/serve/archive/model/ModelConfig.java
@@ -11,18 +11,23 @@
 public class ModelConfig {
     private static final Logger logger = LoggerFactory.getLogger(ModelConfig.class);
 
+    public static final int defaultMinWorkers = 1;
+    public static final int defaultBatchSize = 1;
+    public static final int defaultStartupTimeout = 120; // unit: sec
+    public static final int defaultResponseTimeout = 120; // unit: sec
+
     /** the minimum number of workers of a model */
-    private int minWorkers;
+    private int minWorkers = defaultMinWorkers;
     /** the maximum number of workers of a model */
     private int maxWorkers;
     /** the batch size of a model */
-    private int batchSize;
+    private int batchSize = defaultBatchSize;
     /** the maximum delay in msec of a batch of a model */
     private int maxBatchDelay;
     /** the timeout in sec of a specific model's response. */
-    private int responseTimeout = 120; // unit: sec
+    private int responseTimeout = defaultResponseTimeout;
     /** the timeout in sec of a specific model's startup. */
-    private int startupTimeout = 120; // unit: sec
+    private int startupTimeout = defaultStartupTimeout;
     /**
      * the device type where the model is loaded. It can be gpu, cpu. The model is loaded on CPU if
      * deviceType: "cpu" is set on a GPU host.

diff --git a/frontend/server/src/main/java/org/pytorch/serve/wlm/Model.java b/frontend/server/src/main/java/org/pytorch/serve/wlm/Model.java
@@ -193,32 +193,21 @@ public void setModelState(JsonObject modelInfo) {
         minWorkers =
                 modelInfo.has(MIN_WORKERS) && !modelInfo.get(MIN_WORKERS).isJsonNull()
                         ? modelInfo.get(MIN_WORKERS).getAsInt()
-                        : 1; // default value for minWorkers
-
-        maxWorkers =
-                modelInfo.has(MAX_WORKERS) && !modelInfo.get(MAX_WORKERS).isJsonNull()
-                        ? modelInfo.get(MAX_WORKERS).getAsInt()
-                        : 5; // default value for maxWorkers
-
-        maxBatchDelay =
-                modelInfo.has(MAX_BATCH_DELAY) && !modelInfo.get(MAX_BATCH_DELAY).isJsonNull()
-                        ? modelInfo.get(MAX_BATCH_DELAY).getAsInt()
-                        : 100; // default value for maxBatchDelay
-
+                        : modelArchive.getModelConfig().defaultMinWorkers; // default value for minWorkers
+        maxWorkers = modelInfo.get(MAX_WORKERS).getAsInt();
+        maxBatchDelay = modelInfo.get(MAX_BATCH_DELAY).getAsInt();
         responseTimeout =
                 modelInfo.has(RESPONSE_TIMEOUT) && !modelInfo.get(RESPONSE_TIMEOUT).isJsonNull()
                         ? modelInfo.get(RESPONSE_TIMEOUT).getAsInt()
-                        : 120; // default value for responseTimeout
-
+                        : modelArchive.getModelConfig().defaultResponseTimeout; // default value for responseTimeout
         startupTimeout =
                 modelInfo.has(STARTUP_TIMEOUT) && !modelInfo.get(STARTUP_TIMEOUT).isJsonNull()
                         ? modelInfo.get(STARTUP_TIMEOUT).getAsInt()
-                        : 120; // default value for startupTimeout
-
+                        : modelArchive.getModelConfig().defaultStartupTimeout; // default value for startupTimeout
         batchSize =
                 modelInfo.has(BATCH_SIZE) && !modelInfo.get(BATCH_SIZE).isJsonNull()
                         ? modelInfo.get(BATCH_SIZE).getAsInt()
-                        : 1; // default value for batchSize
+                        : modelArchive.getModelConfig().defaultBatchSize; // default value for batchSize
 
         JsonElement runtime = modelInfo.get(RUNTIME_TYPE);
         String runtime_str = Manifest.RuntimeType.PYTHON.getValue();