wip

.github/workflows/debug-metadata.yml

@@ -1,61 +1,31 @@
 name: Debug Metadata
 
 on:
-  # push:
-  workflow_dispatch:
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
-  cancel-in-progress: true
+  push:
 
 jobs:
-  integration-tests:
-    name: Run TGI Integration Tests
-    runs-on:
-      group: gcp-ct5lp-hightpu-8t
-
+  debug-dind:
+    name: Run TGI tests
+    runs-on: gcp-ct5lp-hightpu-8t-usc1-public-211
+    container:
+      image: us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla@sha256:8f1dcd5b03f993e4da5c20d17c77aff6a5f22d5455f8eb042d2e4b16ac460526
+      options: ${{ vars.V5_LITEPOD_8_ENV }} --shm-size "16gb" --privileged --ipc host -v /mnt/hf_cache:/mnt/cache/
+    env:
+      PJRT_DEVICE: TPU
     steps:
-      - name: Install DNS utilities
+      - name: Install docker
         run: |
-          sudo apt-get update -y
-          sudo apt-get install -y dnsutils iputils-ping curl
-
-    
-
-      - name: Debug metadata
+          apt-get update -y
+          apt-get install -y docker.io
+      - name: Debug XLA (run container)
         run: |
-          echo "nslookup metadata.google.internal"
-          nslookup metadata.google.internal
-
-          echo "ping -c 3 metadata.google.internal"
-          ping -c 3 metadata.google.internal
-
-          echo "ping -c 3 169.254.169.254"
-          ping -c 3 169.254.169.254
-
-          echo "Get the IP address of metadata.google.internal using getent"
-          # Get the IP address of metadata.google.internal using getent
-          METADATA_IP=$(getent hosts metadata.google.internal | awk '{ print $1 }')
-          echo "Metadata server IP address: ${METADATA_IP}"
-          
-          echo "Fetching instance metadata directory:"
-          curl -s http://metadata.google.internal/computeMetadata/v1/instance/ -H "Metadata-Flavor: Google"
-          echo
-
-          STATUS_CODE=$(curl -s -o /dev/null -w "%{http_code}" "http://metadata.google.internal/computeMetadata/v1/instance/image" -H "Metadata-Flavor: Google")
-          RESPONSE=$(curl -s "http://metadata.google.internal/computeMetadata/v1/instance/image" -H "Metadata-Flavor: Google")
-          echo "Status code: $STATUS_CODE"
-          echo "Metadata response: $RESPONSE"
-          if [ "$STATUS_CODE" -ne 200 ]; then
-            echo "Failed to fetch instance image metadata"
-            exit 1
-          fi
-
-          STATUS_CODE=$(curl -s -o /dev/null -w "%{http_code}" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/tpu-env" -H "Metadata-Flavor: Google")
-          RESPONSE=$(curl -s "http://metadata.google.internal/computeMetadata/v1/instance/attributes/tpu-env" -H "Metadata-Flavor: Google")
-          echo "Status code: $STATUS_CODE"
-          echo "Metadata response: $RESPONSE"
-          if [ "$STATUS_CODE" -ne 200 ]; then
-            echo "Failed to fetch TPU environment metadata"
-            exit 1
-          fi
+          docker run \
+            --privileged \
+            --ipc host \
+            --shm-size=16G \
+            -v /mnt/hf_cache:/mnt/cache/ \
+            -e PJRT_DEVICE=TPU \
+            us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla@sha256:8f1dcd5b03f993e4da5c20d17c77aff6a5f22d5455f8eb042d2e4b16ac460526 \
+            bash -c '
+                python -c "import torch_xla.core.xla_model as xm; assert xm.xla_device().type == 'xla', 'XLA device not available'"
+            '

.github/workflows/test-pytorch-xla-tpu-tgi-integration.yml

@@ -1,7 +1,7 @@
 name: Optimum TPU / Test TGI on TPU / Integration Tests
 
 on:
-  push:
+  # push:
   pull_request:
     branches: [ main ]
     paths: