platform framework

.pre-commit-config.yaml

@@ -2,7 +2,7 @@
 # See https://pre-commit.com/hooks.html for more hooks
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.0.1
+    rev: v4.4.0
     hooks:
       - id: check-ast
       - id: check-builtin-literals
@@ -18,23 +18,23 @@ repos:
       - id: check-vcs-permalinks
       - id: check-shebang-scripts-are-executable
   - repo: https://github.com/pre-commit/pygrep-hooks
-    rev: v1.9.0
+    rev: v1.10.0
     hooks:
       - id: python-check-mock-methods
       - id: python-no-log-warn
       - id: python-use-type-annotations
   - repo: https://github.com/hadialqattan/pycln
-    rev: v1.2.5
+    rev: v2.1.3
     hooks:
       - id: pycln
         args: [--all]
   - repo: https://github.com/psf/black
-    rev: 22.3.0
+    rev: 23.1.0
     hooks:
       - id: black
         additional_dependencies: ['click==8.0.4']
   - repo: https://github.com/PyCQA/isort
-    rev: 5.11.5
+    rev: 5.12.0
     hooks:
       - id: isort
         args: ["--profile", "black"]

docs/configuration.md

@@ -266,6 +266,11 @@ models={\
   }\
 }
 ```
+As of version 0.8.0, TorchServe allows for model configuration using a YAML file embedded in the MAR file. This YAML file contains two distinct parts that determine how a model is configured: frontend parameters and backend parameters. (see [details](https://github.com/pytorch/serve/tree/master/model-archiver))
+
+* The frontend parameters are controlled by TorchServe's frontend and specify the parameter name and default values. TorchServe now uses a priority order to determine the final value of a model's parameters in frontend. Specifically, the config.property file has the lowest priority, followed by the model configuration YAML file, and finally, the REST or gRPC model management API has the highest priority.
+
+* The backend parameters are fully controlled by the user. Users customized handler can access the backend parameters via the `model_yaml_config` property of the context object.
 
 ### Other properties
 

frontend/archive/build.gradle

@@ -3,6 +3,7 @@ dependencies {
     api "org.slf4j:slf4j-api:${slf4j_api_version}"
     api "org.apache.logging.log4j:log4j-slf4j-impl:${slf4j_log4j_version}"
     api "com.google.code.gson:gson:${gson_version}"
+    implementation "org.yaml:snakeyaml:${snakeyaml_version}"
 
     testImplementation "commons-cli:commons-cli:${commons_cli_version}"
     testImplementation "org.testng:testng:${testng_version}"

frontend/archive/src/main/java/org/pytorch/serve/archive/model/Manifest.java

@@ -64,6 +64,7 @@ public static final class Model {
         private String handler;
         private String envelope;
         private String requirementsFile;
+        private String configFile;
 
         public Model() {}
 
@@ -122,6 +123,14 @@ public String getEnvelope() {
         public void setEnvelope(String envelope) {
             this.envelope = envelope;
         }
+
+        public String getConfigFile() {
+            return configFile;
+        }
+
+        public void setConfigFile(String configFile) {
+            this.configFile = configFile;
+        }
     }
 
     public enum RuntimeType {

frontend/archive/src/main/java/org/pytorch/serve/archive/model/ModelArchive.java

@@ -6,6 +6,7 @@
 import java.nio.file.FileAlreadyExistsException;
 import java.nio.file.Files;
 import java.util.List;
+import java.util.Map;
 import org.apache.commons.io.FileUtils;
 import org.apache.commons.io.FilenameUtils;
 import org.pytorch.serve.archive.DownloadArchiveException;
@@ -25,12 +26,14 @@ public class ModelArchive {
     private String url;
     private File modelDir;
     private boolean extracted;
+    private ModelConfig modelConfig;
 
     public ModelArchive(Manifest manifest, String url, File modelDir, boolean extracted) {
         this.manifest = manifest;
         this.url = url;
         this.modelDir = modelDir;
         this.extracted = extracted;
+        this.modelConfig = null;
     }
 
     public static ModelArchive downloadModel(
@@ -92,7 +95,7 @@ private static ModelArchive load(String url, File dir, boolean extracted)
         boolean failed = true;
         try {
             File manifestFile = new File(dir, "MAR-INF/" + MANIFEST_FILE);
-            Manifest manifest = null;
+            Manifest manifest;
             if (manifestFile.exists()) {
                 manifest = ArchiveUtils.readFile(manifestFile, Manifest.class);
             } else {
@@ -179,4 +182,21 @@ public void clean() {
             FileUtils.deleteQuietly(modelDir);
         }
     }
+
+    public ModelConfig getModelConfig() {
+        if (this.modelConfig == null && manifest.getModel().getConfigFile() != null) {
+            try {
+                File configFile =
+                        new File(modelDir.getAbsolutePath(), manifest.getModel().getConfigFile());
+                Map<String, Object> modelConfigMap = ArchiveUtils.readYamlFile(configFile);
+                this.modelConfig = ModelConfig.build(modelConfigMap);
+            } catch (InvalidModelException | IOException e) {
+                logger.error(
+                        "Failed to parse model config file {}",
+                        manifest.getModel().getConfigFile(),
+                        e);
+            }
+        }
+        return this.modelConfig;
+    }
 }