Merge pull request #143 from JoeyHwong-gk/lldoc

improve lifelong learning docs

build/crd-samples/sedna/lifelonglearningjobv1alpha1.yaml

@@ -0,0 +1,80 @@
+kubectl create -f - <<EOF
+apiVersion: sedna.io/v1alpha1
+kind: LifelongLearningJob
+metadata:
+  name: atcii-classifier-demo
+spec:
+  dataset:
+    name: "lifelong-dataset"
+    trainProb: 0.8
+  trainSpec:
+    template:
+      spec:
+        nodeName:  "edge-node"
+        containers:
+          - image: kubeedge/sedna-example-lifelong-learning-atcii-classifier:v0.3.0
+            name:  train-worker
+            imagePullPolicy: IfNotPresent
+            args: ["train.py"]
+            env:
+              - name: "early_stopping_rounds"
+                value: "100"
+              - name: "metric_name"
+                value: "mlogloss"
+    trigger:
+      checkPeriodSeconds: 60
+      timer:
+        start: 02:00
+        end: 24:00
+      condition:
+        operator: ">"
+        threshold: 500
+        metric: num_of_samples
+  evalSpec:
+    template:
+      spec:
+        nodeName:  "edge-node"
+        containers:
+          - image: kubeedge/sedna-example-lifelong-learning-atcii-classifier:v0.3.0
+            name:  eval-worker
+            imagePullPolicy: IfNotPresent
+            args: ["eval.py"]
+            env:
+              - name: "metrics"
+                value: "precision_score"
+              - name: "metric_param"
+                value: "{'average': 'micro'}"
+              - name: "model_threshold"
+                value: "0.5"
+  deploySpec:
+    template:
+      spec:
+        nodeName:  "edge-node"
+        containers:
+        - image: kubeedge/sedna-example-lifelong-learning-atcii-classifier:v0.3.0
+          name:  infer-worker
+          imagePullPolicy: IfNotPresent
+          args: ["inference.py"]
+          env:
+          - name: "UT_SAVED_URL"
+            value: "/ut_saved_url"
+          - name: "infer_dataset_url"
+            value: "/data/testData.csv"
+          volumeMounts:
+          - name: utdir
+            mountPath: /ut_saved_url
+          - name: inferdata
+            mountPath: /data/
+          resources:
+            limits:
+              memory: 2Gi
+        volumes:
+          - name: utdir
+            hostPath:
+              path: /lifelong/unseen_task/
+              type: DirectoryOrCreate
+          - name: inferdata
+            hostPath:
+              path:  /data/
+              type: DirectoryOrCreate
+  outputDir: "/output"

docs/proposals/lifelong-learning.md

@@ -65,11 +65,11 @@ The tables below summarize the group, kind and API version details for the CRD.
 |Kind                   | LifelongLearningJob             |
 
 ### Lifelong learning CRD
-See the [crd source](/build/crds/sedna/Lifelonglearningjob_v1alpha1.yaml) for details.
+See the [crd source](/build/crds/sedna/sedna.io_lifelonglearningjobs.yaml) for details.
 
 ### Lifelong learning job type definition
 
-See the [golang source](/pkg/apis/sedna/v1alpha1/Lifelongllearningjob_types.go) for details.
+See the [golang source](/pkg/apis/sedna/v1alpha1/lifelonglearningjob_types.go) for details.
 
 #### Validation
 [Open API v3 Schema based validation](https://kubernetes.io/docs/tasks/access-kubernetes-api/custom-resources/custom-resource-definitions/#validation) can be used to guard against bad requests.
@@ -80,7 +80,7 @@ Here is a list of validations we need to support :
 1. The edgenode name specified in the crd should exist in k8s.
 
 ### Lifelong learning job sample
-See the [source](/build/crd-samples/sedna/Lifelonglearningjob_v1alpha1.yaml) for an example.
+See the [source](/build/crd-samples/sedna/lifelonglearningjob_v1alpha1.yaml) for an example.
     
 ## Controller Design
 

examples/lifelong_learning/atcii/README.md

@@ -1,8 +1,8 @@
 # Using Lifelong Learning Job in Thermal Comfort Prediction Scenario
 
 This document introduces how to use lifelong learning job in thermal comfort prediction scenario. 
-Using the lifelong learning job, our application can automatically retrains, evaluates, 
-and updates models based on the data generated at the edge.
+Using the lifelong learning job, our application can automatically retrain, evaluate, 
+and update models based on the data generated at the edge.
 
 ##  Thermal Comfort Prediction Experiment
 
@@ -16,15 +16,15 @@ In this example, you can use [ASHRAE Global Thermal Comfort Database II](https:/
 
 
 
-download [datasets](https://kubeedge.obs.cn-north-1.myhuaweicloud.com/examples/atcii-classifier/dataset.tar.gz), including train, evaluation and incremental dataset.
+We provide a well-processed [datasets](https://kubeedge.obs.cn-north-1.myhuaweicloud.com/examples/atcii-classifier/dataset.tar.gz), including train (`trainData.csv`), evaluation (`testData.csv`) and incremental (`trainData2.csv`) dataset.
 ```
-cd /
+cd /data
 wget https://kubeedge.obs.cn-north-1.myhuaweicloud.com/examples/atcii-classifier/dataset.tar.gz
 tar -zxvf dataset.tar.gz
 ```
 
 ### Create Lifelong Job
-in this example, `$WORKER_NODE` is a custom node, you can fill it which you actually run.
+In this example, `$WORKER_NODE` is a custom node, you can fill it which you actually run.
 
 
 ```
@@ -39,13 +39,13 @@ kind: Dataset
 metadata:
   name: lifelong-dataset
 spec:
-  url: "/data/lifelong_learning/trainData.csv"
+  url: "/data/trainData.csv"
   format: "csv"
   nodeName: $WORKER_NODE
 EOF
 ```
 
-Also, you can trigger retraining by use `incremental Dataset`[trainData2.csv] to replace `trainData.csv` 
+Also, you can replace `trainData.csv` with `trainData2.csv` which contained in `dataset` to trigger retraining.
 
 Start The Lifelong Learning Job
 
@@ -68,8 +68,8 @@ spec:
           - image: kubeedge/sedna-example-lifelong-learning-atcii-classifier:v0.3.0
             name:  train-worker
             imagePullPolicy: IfNotPresent
-            args: ["train.py"]
-            env:
+            args: ["train.py"]  # training script
+            env:  # Hyperparameters required for training
               - name: "early_stopping_rounds"
                 value: "100"
               - name: "metric_name"
@@ -97,7 +97,7 @@ spec:
                 value: "precision_score"
               - name: "metric_param"
                 value: "{'average': 'micro'}"
-              - name: "model_threshold"
+              - name: "model_threshold"  # Threshold for filtering deploy models
                 value: "0.5"
   deploySpec:
     template:
@@ -109,9 +109,9 @@ spec:
           imagePullPolicy: IfNotPresent
           args: ["inference.py"]
           env:
-          - name: "UT_SAVED_URL"
+          - name: "UT_SAVED_URL"  # unseen tasks save path
             value: "/ut_saved_url"
-          - name: "infer_dataset_url"
+          - name: "infer_dataset_url"  # simulation of the inference samples 
             value: "/data/testData.csv"
           volumeMounts:
           - name: utdir
@@ -134,6 +134,8 @@ spec:
 EOF
 ```
 
+>**Note**: `outputDir` can be set as s3 storage url to save artifacts(model, sample, etc.) into s3, and follow [this](/examples/storage/s3/README.md) to set the credentials.
+
 ### Check Lifelong Learning Job
 query the service status
 ```