Predict registry own model.

.gitignore

@@ -7,4 +7,5 @@ plot.png
 my_runs
 mlruns
 my_model
-1/
+1/
+mydb.sqlite

Jenkinsfile_predict

@@ -1,9 +1,5 @@
 pipeline {
 	agent any 
-//	triggers {
-//		upstream(upstreamProjects: "s426206-train/master", 
-//		threshold: hudson.model.Result.SUCCESS) 	
-//	}
     parameters {
       buildSelector(
           defaultSelector: lastSuccessful(),

Jenkinsfile_predict_registry

@@ -0,0 +1,26 @@
+pipeline {
+	agent any
+	stages {
+		stage('checkout') {
+			steps {
+				checkout([$class: 'GitSCM', branches: [[name: '*/master']], doGenerateSubmoduleConfigurations: false, extensions: [], submoduleCfg: [], userRemoteConfigs: [[url: 'https://git.wmi.amu.edu.pl/s426206/ium_426206.git']]])
+			}
+		}	
+        stage('Copy artifact') {
+            steps {
+                copyArtifacts filter: 'my_model/input_example.json', fingerprintArtifacts: false, projectName: 's426206-training/master', selector: buildParameter('BUILD_SELECTOR_PREDICT')
+            }
+        }
+		stage('docker') {
+			steps {
+                script {
+                    def img = docker.build('rokoch/ium:01')
+                    img.inside {
+                        sh 'chmod +x mlflow_predict_registry.py'
+                        sh 'python3 ./mlflow_predict_registry.py'
+                    }
+                }
+			}
+		}
+	}
+}

mlflow_predict_registry.py

@@ -0,0 +1,24 @@
+import mlflow
+import mlflow.pytorch
+from mlflow.tracking import MlflowClient
+import numpy as np
+import torch
+import json
+
+#mlflow.set_tracking_uri("http://127.0.0.1:5000")
+mlflow.set_tracking_uri("http://172.17.0.1:5000")
+client = MlflowClient()
+version = 0
+model_name = "s426206"
+for mv in client.search_model_versions("name='s426206'"):
+    if int(mv.version) > version:
+        version = int(mv.version)
+
+model = mlflow.pytorch.load_model(
+    model_uri=f"models:/{model_name}/{version}"
+)
+
+with open('my_model/input_example.json') as json_file:
+    data = json.load(json_file)
+#print(np.array(data['inputs']))
+print(model(torch.tensor(np.array(data['inputs'])).float()))