Refactor build-and-run workflow with model matrix

Updated workflow to include matrix strategy for models and modified test prompt.

Author: mikepapadim

.github/workflows/build-and-run.yml

@@ -1,5 +1,4 @@
 name: GPULlama3 Build & Run
-
 on:
   push:
     branches: [ main ]
@@ -8,20 +7,29 @@ on:
     types: [opened, synchronize, reopened]
 
 jobs:
-  build-and-run:
+  build-and-test:
     runs-on: self-hosted
 
+    strategy:
+      fail-fast: false
+      matrix:
+        model:
+          - Llama-3.2-1B-Instruct-F16.gguf
+          - Qwen3-4B-f16.gguf
+          - DeepSeek-R1-Distill-Qwen-1.5B-F16.gguf
+    
     env:
       JAVA_HOME: /opt/jenkins/jdks/graal-23.1.0/jdk-21.0.3
       TORNADO_ROOT: ${{ github.workspace }}/GPULlama3.java/external/tornadovm
-      LLAMA_ROOT: ${{ github.workspace }} 
+      LLAMA_ROOT: ${{ github.workspace }}
+      MODEL_DIR: /home/michalis/models
 
     steps:
       - name: Checkout GPULlama3
         uses: actions/checkout@v4
         with:
           fetch-depth: 0
-
+          
       - name: Check code formatting (Spotless)
         run: |
           cd ${{ github.workspace }}
@@ -32,17 +40,12 @@ jobs:
           git clone --depth 1 --branch master \
             https://github.com/beehive-lab/TornadoVM.git \
             GPULlama3.java/external/tornadovm
-
-      - name: Set up Python venv for TornadoVM
-        run: |
-          python3 -m venv GPULlama3.java/external/tornadovm/venv
-          source GPULlama3.java/external/tornadovm/venv/bin/activate
-          python --version
-
+            
       - name: Build TornadoVM
         run: |
           set -x
           cd GPULlama3.java/external/tornadovm
+          python3 -m venv venv
           source venv/bin/activate
           echo "=== Building TornadoVM ==="
           make
@@ -56,18 +59,16 @@ jobs:
           FULL_SDK="${PWD}/${SDK_DIR}"
           echo "Detected TornadoVM SDK: $FULL_SDK"
           
-          # Export for current shell session
           export TORNADO_SDK="$FULL_SDK"
           export PATH="$FULL_SDK/bin:$JAVA_HOME/bin:$PATH"
           
-          # Save for subsequent steps
           echo "TORNADO_SDK=$FULL_SDK" >> $GITHUB_ENV
           echo "PATH=$PATH" >> $GITHUB_ENV
           
           echo "=== Checking tornado CLI ==="
           which tornado || { echo "::error::tornado not in PATH"; exit 1; }
           tornado --devices
-
+          
       - name: Build GPULlama3
         run: |
           set -x
@@ -77,13 +78,16 @@ jobs:
           which tornado || { echo "::error::tornado unavailable during GPULlama3 build"; exit 1; }
           tornado --version
           make
-
-      - name: Run llama-tornado test prompt
+          
+      - name: Test Inference ${{ matrix.model.name }} with OpenCL
         run: |
           set -x
           cd ${{ github.workspace }}
           export PATH="$TORNADO_SDK/bin:$JAVA_HOME/bin:$PATH"
           which tornado || { echo "::error::tornado not found at runtime"; exit 1; }
+          
+          echo "=== Testing ${{ matrix.model.name }} ==="
           ./llama-tornado --gpu --opencl \
-            --model /home/michalis/models/Llama-3.2-1B-Instruct-F16.gguf \
-            --prompt "Say hello"
+            --model $MODEL_DIR/${{ matrix.model.file }} \
+            --prompt "Tell me a joke" \
+            --max-tokens 50