Fix test workflows to use build artifacts instead of local binary

- Build workflow now uploads ollama binary as artifact with 7-day retention
- Test workflows download artifact instead of expecting local binary
- Eliminates 'ollama binary not found' error when running tests
- Enables build-once, test-multiple-times workflow pattern
- Added binary verification step to confirm artifact download

.github/workflows/tesla-k80-multi-gpu-tests.yml

@@ -4,7 +4,7 @@ on:
   workflow_dispatch: # Manual trigger only
   schedule:
     # Run weekly on Sundays at 2 AM UTC (less frequent than single-GPU tests)
-    - cron: '0 2 * * 0'
+    - cron: "0 2 * * 0"
 
 jobs:
   multi-gpu-test:
@@ -16,13 +16,17 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v4
 
-      - name: Verify ollama binary exists
+      - name: Download ollama binary from build workflow
+        uses: actions/download-artifact@v4
+        with:
+          name: ollama-binary
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Make ollama binary executable
         run: |
-          if [ ! -f ./ollama ]; then
-            echo "Error: ollama binary not found. Please run the build workflow first."
-            exit 1
-          fi
+          chmod +x ollama
           ls -lh ollama
+          ./ollama --version
 
       - name: Verify multi-GPU setup
         run: |