fix for snapshot variables missing/null (#3328)

* testing adding startup * testing adding default * testing newer docker * adding default values for model * testing specific docker * testing specific docker image * fixed format * merge ready * adding source of truth * adding source of truth * fixing tests * format * removing defaults --------- Co-authored-by: Ankith Gunapal <agunapal@ischool.Berkeley.edu>
pytorch · Sep 23, 2024 · c161926 · c161926
1 parent 6881ec5
commit c161926
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 4 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
@@ -11,6 +11,9 @@
 public class ModelConfig {
     private static final Logger logger = LoggerFactory.getLogger(ModelConfig.class);
 
+    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;
     /** the maximum number of workers of a model */
@@ -20,9 +23,9 @@ public class ModelConfig {
     /** 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
@@ -193,9 +193,17 @@ public void setModelState(JsonObject modelInfo) {
         minWorkers = modelInfo.get(MIN_WORKERS).getAsInt();
         maxWorkers = modelInfo.get(MAX_WORKERS).getAsInt();
         maxBatchDelay = modelInfo.get(MAX_BATCH_DELAY).getAsInt();
-        responseTimeout = modelInfo.get(RESPONSE_TIMEOUT).getAsInt();
-        startupTimeout = modelInfo.get(STARTUP_TIMEOUT).getAsInt();
         batchSize = modelInfo.get(BATCH_SIZE).getAsInt();
+        responseTimeout =
+                modelInfo.has(RESPONSE_TIMEOUT) && !modelInfo.get(RESPONSE_TIMEOUT).isJsonNull()
+                        ? modelInfo.get(RESPONSE_TIMEOUT).getAsInt()
+                        : modelArchive.getModelConfig()
+                                .defaultResponseTimeout; // default value for responseTimeout
+        startupTimeout =
+                modelInfo.has(STARTUP_TIMEOUT) && !modelInfo.get(STARTUP_TIMEOUT).isJsonNull()
+                        ? modelInfo.get(STARTUP_TIMEOUT).getAsInt()
+                        : modelArchive.getModelConfig()
+                                .defaultStartupTimeout; // default value for startupTimeout
 
         JsonElement runtime = modelInfo.get(RUNTIME_TYPE);
         String runtime_str = Manifest.RuntimeType.PYTHON.getValue();