Merge branch 'dev' into main

rh-aiservices-bu · Apr 9, 2024 · 7caf9a5 · 7caf9a5
2 parents 855e541 + 0b3faa5
commit 7caf9a5
Show file tree

Hide file tree

Showing 80 changed files with 2,119 additions and 2,193 deletions.
diff --git a/.github/.wordlist.txt b/.github/.wordlist.txt
@@ -49,6 +49,8 @@ DevOps
 dex
 diag
 dir
+diy
+DIY
 dotenv
 drawio
 ds
@@ -107,6 +109,7 @@ Inlining
 insecureEdgeTerminationPolicy
 io
 ipynb
+jpg
 js
 json
 jsonpath

diff --git a/bootstrap/ic-rhoai-configuration/images-puller.yaml b/bootstrap/ic-rhoai-configuration/images-puller.yaml
@@ -19,7 +19,7 @@ spec:
     spec:
       containers:
       - name: ic-workbench
-        image: image-registry.openshift-image-registry.svc:5000/redhat-ods-applications/ic-workbench:1.2
+        image: image-registry.openshift-image-registry.svc:5000/redhat-ods-applications/ic-workbench:2.0
         command: ["tail"]
         args: ["-f", "/dev/null"]
         resources:

diff --git a/bootstrap/ic-rhoai-configuration/odhdashboardconfig.yaml b/bootstrap/ic-rhoai-configuration/odhdashboardconfig.yaml
@@ -19,7 +19,6 @@ spec:
     enabled: true
     notebookNamespace: rhods-notebooks
     pvcSize: 5Gi
-    gpuSetting: '4'
   notebookSizes:
     - name: Standard
       resources:

diff --git a/bootstrap/ic-rhoai-configuration/workbench-imagestream.yaml b/bootstrap/ic-rhoai-configuration/workbench-imagestream.yaml
@@ -25,17 +25,17 @@ spec:
   lookupPolicy:
     local: true
   tags:
-    - name: '1.2'
+    - name: '2.0'
       annotations:
         opendatahub.io/notebook-python-dependencies: >-
-          [{"name":"PyTorch","version":"2.1.2"},{"name":"Langchain","version":"0.0.353"},{"name":"Ultralytics","version":"8.0.232"},]
+          [{"name":"PyTorch","version":"2.1.2"},{"name":"Langchain","version":"0.1.12"},{"name":"Ultralytics","version":"8.1.29"},]
         opendatahub.io/notebook-software: >-
           [{"name":"CUDA","version":"12.1"},{"name":"Python","version":"v3.11"}]
         openshift.io/imported-from: quay.io/rh-aiservices-bu/rhoai-lab-insurance-claim-workbench
       from:
         kind: DockerImage
         name: >-
-          quay.io/rh-aiservices-bu/rhoai-lab-insurance-claim-workbench:1.2
+          quay.io/rh-aiservices-bu/rhoai-lab-insurance-claim-workbench:2.0
       importPolicy:
         importMode: Legacy
       referencePolicy:

diff --git a/bootstrap/ic-shared-llm/deployment-flant5.yaml b/bootstrap/ic-shared-llm/deployment-flant5.yaml
diff --git a/bootstrap/ic-shared-llm/deployment-ollama.yaml b/bootstrap/ic-shared-llm/deployment-ollama.yaml
@@ -0,0 +1,81 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ollama
+  namespace: ic-shared-llm
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: ollama
+  strategy:
+    type: Recreate
+  template:
+    metadata:
+      labels:
+        app: ollama
+    spec:
+      containers:
+      - name: ollama
+        image: quay.io/rh-aiservices-bu/ollama-ubi9:0.1.30
+        imagePullPolicy: Always
+        env:
+        - name: MODEL_ID
+          value: mistral:latest
+        - name: OLLAMA_MODELS
+          value: /.ollama/models
+        - name: OLLAMA_HOST
+          value: 0.0.0.0
+        - name: OLLAMA_KEEP_ALIVE
+          value: '-1m'
+        securityContext:
+            capabilities:
+              drop:
+                - ALL
+            runAsNonRoot: true
+            allowPrivilegeEscalation: false
+            seccompProfile:
+              type: RuntimeDefault
+        ports:
+        - name: http
+          containerPort: 11434
+          protocol: TCP
+        resources:
+          limits:
+            cpu: '4'
+            memory: 8Gi
+          requests:
+            cpu: '4'
+            memory: 8Gi
+        readinessProbe:
+          httpGet:
+            path: /
+            port: http
+            scheme: HTTP
+          timeoutSeconds: 5
+          periodSeconds: 30
+          successThreshold: 1
+          failureThreshold: 3
+        livenessProbe:
+          httpGet:
+            path: /
+            port: http
+            scheme: HTTP
+          timeoutSeconds: 8
+          periodSeconds: 100
+          successThreshold: 1
+          failureThreshold: 3
+        lifecycle:
+            postStart:
+              exec:
+                command: [ "/bin/sh", "-c", "/bin/ollama pull $MODEL_ID" ]
+        terminationMessagePath: /dev/termination-log
+        terminationMessagePolicy: File
+        volumeMounts:
+        - mountPath: /.ollama
+          name: ollama-cache
+      restartPolicy: Always
+      volumes:
+      - name: ollama-cache
+        persistentVolumeClaim:
+          claimName: ollama-cache
diff --git a/bootstrap/ic-shared-llm/deployment.yaml b/bootstrap/ic-shared-llm/deployment.yaml
@@ -49,19 +49,14 @@ spec:
             periodSeconds: 100
             successThreshold: 1
             failureThreshold: 3
-          env:
-            - name: MODEL_ID
-              value: mistralai/Mistral-7B-Instruct-v0.2
-            - name: MAX_INPUT_LENGTH
-              value: '4096'
-            - name: MAX_TOTAL_TOKENS
-              value: '8192'
-            - name: HUGGINGFACE_HUB_CACHE
-              value: /models-cache
-            - name: PORT
-              value: '3000'
-            - name: HOSTNAME
-              value: '0.0.0.0'
+          env: []
+          args: [
+            "--model",
+            "mistralai/Mistral-7B-Instruct-v0.2",
+            "--download-dir",
+            "/models-cache",
+            "--dtype", "float16",
+            "--max-model-len", "6144" ]
           securityContext:
             capabilities:
               drop:
@@ -72,7 +67,7 @@ spec:
               type: RuntimeDefault
           ports:
             - name: http
-              containerPort: 3000
+              containerPort: 8000
               protocol: TCP
           imagePullPolicy: IfNotPresent
           startupProbe:
@@ -91,7 +86,7 @@ spec:
             - name: shm
               mountPath: /dev/shm
           terminationMessagePolicy: File
-          image: 'ghcr.io/huggingface/text-generation-inference:1.3'
+          image: 'quay.io/rh-aiservices-bu/vllm-openai-ubi9:0.4.0'
       volumes:
         - name: models-cache
           persistentVolumeClaim:

diff --git a/bootstrap/ic-shared-llm/kustomization.yaml b/bootstrap/ic-shared-llm/kustomization.yaml
@@ -11,8 +11,9 @@ resources:
 - fix-odf-config.yaml
 # wave 1
 - pvc.yaml
+- pvc-ollama.yaml
 - deployment.yaml
 - service.yaml
-- deployment-flant5.yaml
-- service-flant5.yaml
+- deployment-ollama.yaml
+- service-ollama.yaml
 # wave 2
diff --git a/bootstrap/ic-shared-llm/pvc-ollama.yaml b/bootstrap/ic-shared-llm/pvc-ollama.yaml
@@ -0,0 +1,18 @@
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: ollama-cache
+  namespace: ic-shared-llm
+  labels:
+    app: ollama
+  annotations:
+    argocd.argoproj.io/sync-wave: "0"
+spec:
+  accessModes:
+    - ReadWriteMany
+  resources:
+    requests:
+      storage: 10Gi
+  storageClassName: ocs-storagecluster-cephfs
+  volumeMode: Filesystem
diff --git a/bootstrap/ic-shared-llm/service-flant5.yaml → bootstrap/ic-shared-llm/service-ollama.yaml b/bootstrap/ic-shared-llm/service-flant5.yaml → bootstrap/ic-shared-llm/service-ollama.yaml
@@ -1,21 +1,21 @@
 kind: Service
 apiVersion: v1
 metadata:
-  name: llm-flant5
+  name: ollama
   namespace: ic-shared-llm
   labels:
-    app: llm-flant5
+    app: ollama
 spec:
   clusterIP: None
   ipFamilies:
     - IPv4
   ports:
     - name: http
       protocol: TCP
-      port: 3000
+      port: 11434
       targetPort: http
   type: ClusterIP
   ipFamilyPolicy: SingleStack
   sessionAffinity: None
   selector:
-    app: llm-flant5
+    app: ollama
diff --git a/bootstrap/ic-shared-llm/service.yaml b/bootstrap/ic-shared-llm/service.yaml
@@ -12,7 +12,7 @@ spec:
   ports:
     - name: http
       protocol: TCP
-      port: 3000
+      port: 8000
       targetPort: http
   type: ClusterIP
   ipFamilyPolicy: SingleStack
-Original file line number
+Diff line change
@@ Expand Up / @@ -49,6 +49,8 @@ DevOps @@
     dex
     diag
     dir
+    diy
+    DIY
     dotenv
     drawio
     ds
@@ Expand Down Expand Up / @@ -107,6 +109,7 @@ Inlining @@
     insecureEdgeTerminationPolicy
     io
     ipynb
+    jpg
     js
     json
     jsonpath
@@ Expand Down @@