Merge branch 'main' into Ishaan/tuning-oom-prevention

Azure · Oct 18, 2024 · 1b341f6 · 1b341f6
2 parents 4a191d4 + b6694c2
commit 1b341f6
Show file tree

Hide file tree

Showing 161 changed files with 7,945 additions and 1,332 deletions.
diff --git a/.github/e2e-preset-configs.json b/.github/e2e-preset-configs.json
@@ -6,105 +6,129 @@
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
+      },
+      {
+        "name": "falcon-7b-adapter",
+        "node-count": 1,
+        "node-vm-size": "Standard_NC12s_v3",
+        "node-osdisk-size": 100,
+        "OSS": true,
+        "loads_adapter": true,
+        "expected_adapter": "amod-mental-health"
       },
       {
         "name": "falcon-7b-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "falcon-40b",
         "node-count": 1,
         "node-vm-size": "Standard_NC24s_v3",
         "node-osdisk-size": 400,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "falcon-40b-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC24s_v3",
         "node-osdisk-size": 400,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "mistral-7b",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "mistral-7b-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "phi-2",
         "node-count": 1,
         "node-vm-size": "Standard_NC6s_v3",
         "node-osdisk-size": 50,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "phi-3-mini-4k-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC6s_v3",
         "node-osdisk-size": 50,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "phi-3-mini-128k-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC6s_v3",
         "node-osdisk-size": 50,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "phi-3-medium-4k-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "phi-3-medium-128k-instruct",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": true
+        "OSS": true,
+        "loads_adapter": false
       },
       {
         "name": "llama-2-7b",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": false
+        "OSS": false,
+        "loads_adapter": false
       },
       {
         "name": "llama-2-7b-chat",
         "node-count": 1,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 100,
-        "OSS": false
+        "OSS": false,
+        "loads_adapter": false
       },
       {
         "name": "llama-2-13b",
         "node-count": 2,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 150,
-        "OSS": false
+        "OSS": false,
+        "loads_adapter": false
       },
       {
         "name": "llama-2-13b-chat",
         "node-count": 2,
         "node-vm-size": "Standard_NC12s_v3",
         "node-osdisk-size": 150,
-        "OSS": false
+        "OSS": false,
+        "loads_adapter": false
       }
     ]
   }

diff --git a/.github/workflows/codeql.yml b/.github/workflows/codeql.yml
@@ -22,7 +22,7 @@ jobs:
 
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@17d0e2bd7d51742c71671bd19fa12bdc9d40a3d6 # v2.8.1
+        uses: step-security/harden-runner@5c7944e73c4c2a096b17a9cb74d65b6c2bbafbde # v2.9.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/create-release.yml b/.github/workflows/create-release.yml
@@ -16,7 +16,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@17d0e2bd7d51742c71671bd19fa12bdc9d40a3d6 # v2.8.1
+        uses: step-security/harden-runner@5c7944e73c4c2a096b17a9cb74d65b6c2bbafbde # v2.9.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/dependency-review.yml b/.github/workflows/dependency-review.yml
@@ -17,7 +17,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@17d0e2bd7d51742c71671bd19fa12bdc9d40a3d6 # v2.8.1
+        uses: step-security/harden-runner@5c7944e73c4c2a096b17a9cb74d65b6c2bbafbde # v2.9.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/e2e-preset-test.yml b/.github/workflows/e2e-preset-test.yml
@@ -119,7 +119,7 @@ jobs:
         run: echo "MODEL_IS_OSS=${{ matrix.model.OSS }}" >> $GITHUB_ENV
 
       - name: 'Az CLI login'
-        uses: azure/login@v2.1.1
+        uses: azure/login@v2.2.0
         with:
             client-id: ${{ secrets.AZURE_CLIENT_ID }}
             tenant-id: ${{ secrets.AZURE_TENANT_ID }}
@@ -237,7 +237,13 @@ jobs:
       - name: Wait for Resource to be ready
         run: |
             kubectl rollout status ${{steps.resource.outputs.RESOURCE_TYPE}}/${{ matrix.model.name }} --timeout=1800s
-        
+    
+      - name: Check Adapter Loading from Logs
+        if: matrix.model.loads_adapter == true
+        run: |
+            POD_NAME=$(kubectl get pods -l app=${{ matrix.model.name }} -o jsonpath="{.items[0].metadata.name}")
+            kubectl logs $POD_NAME | grep "Adapter added:" | grep "${{ matrix.model.expected_adapter }}" || (echo "Adapter not loaded or incorrect adapter loaded" && exit 1)
+          
       - name: Test home endpoint
         run: |
             curl http://${{ steps.get_ip.outputs.SERVICE_IP }}:80/