Path updates and notebook testing

.github/workflows/test-notebooks.yml

@@ -0,0 +1,127 @@
+name: Test Notebooks
+
+on:
+  push:
+    branches: [ main, develop ]
+  pull_request:
+    branches: [ main ]
+  schedule:
+    # Run tests weekly to catch SDK updates
+    - cron: '0 6 * * 1'  # Monday at 6 AM UTC
+
+jobs:
+  test-notebooks:
+    runs-on: ubuntu-latest
+
+    # This job runs on Trainium instances with Neuron SDK
+
+    strategy:
+      matrix:
+        notebook-group:
+          - "NxD"
+          - "FineTuning" 
+          - "vLLM"
+          - "NKI"
+      fail-fast: false
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Set up Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: '3.10'
+
+    - name: Install test dependencies
+      run: |
+        pip install -r requirements-test.txt
+
+    - name: Validate notebook syntax (without execution)
+      run: |
+        # Just validate that notebooks are well-formed JSON
+        python -c "
+        import json
+        from pathlib import Path
+
+        notebooks = list(Path('labs/${{ matrix.notebook-group }}').rglob('*.ipynb'))
+        print(f'Validating {len(notebooks)} notebooks in ${{ matrix.notebook-group }}')
+
+        for nb_path in notebooks:
+            if '.ipynb_checkpoints' in str(nb_path):
+                continue
+            try:
+                with open(nb_path) as f:
+                    json.load(f)
+                print(f'✓ {nb_path}')
+            except Exception as e:
+                print(f'✗ {nb_path}: {e}')
+                exit(1)
+        "
+
+    - name: Check for required files
+      run: |
+        # Verify that notebooks reference existing files
+        python -c "
+        import json
+        from pathlib import Path
+        import re
+
+        def check_notebook_references(nb_path):
+            with open(nb_path) as f:
+                nb = json.load(f)
+
+            issues = []
+            for cell in nb.get('cells', []):
+                if cell.get('cell_type') == 'code':
+                    source = ''.join(cell.get('source', []))
+
+                    # Check for file references
+                    file_refs = re.findall(r'[\"\']([\w\./\-]+\.(?:txt|py|json|yaml|yml|sh))[\"\'']', source)
+                    for file_ref in file_refs:
+                        if not file_ref.startswith('/') and not Path(nb_path.parent / file_ref).exists():
+                            issues.append(f'Missing file reference: {file_ref}')
+
+            return issues
+
+        notebooks = list(Path('labs/${{ matrix.notebook-group }}').rglob('*.ipynb'))
+        all_issues = []
+
+        for nb_path in notebooks:
+            if '.ipynb_checkpoints' in str(nb_path):
+                continue
+            issues = check_notebook_references(nb_path)
+            if issues:
+                all_issues.extend([f'{nb_path}: {issue}' for issue in issues])
+
+        if all_issues:
+            print('Found issues:')
+            for issue in all_issues:
+                print(f'  {issue}')
+            # Don't fail on missing files for now, just warn
+            # exit(1)
+        else:
+            print('No file reference issues found')
+        "
+
+  # Full notebook execution on Trainium instances
+  test-execution:
+    runs-on: self-hosted  # Trainium instance with Neuron SDK
+    needs: test-notebooks
+    if: github.event_name == 'push' || github.event_name == 'schedule'
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Run notebook tests
+      run: |
+        # Run on actual Trainium hardware with Neuron SDK
+        ./run_tests.sh --fast --html-report
+
+    - name: Upload test results
+      uses: actions/upload-artifact@v3
+      if: always()
+      with:
+        name: test-results-${{ matrix.notebook-group }}
+        path: test_report.html

labs/FineTuning/HuggingFaceExample/01_finetuning/FT-Qwen3-1.7B-chess.ipynb

@@ -54,7 +54,7 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "%cd /home/ubuntu/environment/FineTuning/HuggingFaceExample/01_finetuning/assets\n",
+        "%cd /home/ubuntu/build-on-trainium-workshop/labs/FineTuning/HuggingFaceExample/01_finetuning/assets\n",
         "%pip install -q -r requirements.txt\n"
       ]
     },
@@ -95,7 +95,7 @@
         "  finetune_chess_model.py \\\n",
         "  --model_id Qwen/Qwen3-1.7B \\\n",
         "  --tokenizer_id Qwen/Qwen3-1.7B \\\n",
-        "  --output_dir ~/environment/ml/qwen-chess \\\n",
+        "  --output_dir ~/ml/qwen-chess \\\n",
         "  --bf16 True \\\n",
         "  --gradient_checkpointing True \\\n",
         "  --gradient_accumulation_steps 1 \\\n",
@@ -128,11 +128,11 @@
       "outputs": [],
       "source": [
         "!optimum-cli export neuron \\\n",
-        "  --model /home/ubuntu/environment/ml/qwen-chess/merged_model \\\n",
+        "  --model ~/ml/qwen-chess/merged_model \\\n",
         "  --task text-generation \\\n",
         "  --sequence_length 2048 \\\n",
         "  --batch_size 4 \\\n",
-        "  /home/ubuntu/environment/ml/qwen-chess/compiled_model\n"
+        "  ~/ml/qwen-chess/compiled_model\n"
       ]
     },
     {
@@ -163,7 +163,7 @@
         "from vllm import LLM, SamplingParams\n",
         "\n",
         "llm = LLM(\n",
-        "    model=\"/home/ubuntu/environment/ml/qwen-chess/compiled_model\", #local compiled model\n",
+        "    model=\"/home/ubuntu/ml/qwen-chess/compiled_model\", #full path to local compiled model\n",
         "    max_num_seqs=4,\n",
         "    max_model_len=2048,\n",
         "    device=\"neuron\",\n",

labs/FineTuning/HuggingFaceExample/01_finetuning/Finetune-Qwen3-1.7B.ipynb

@@ -38,7 +38,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%cd /home/ubuntu/environment/FineTuning/HuggingFaceExample/01_finetuning/assets\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/FineTuning/HuggingFaceExample/01_finetuning/assets\n",
     "%pip install -r requirements.txt"
    ]
   },
@@ -79,7 +79,7 @@
     "  finetune_model.py \\\n",
     "  --model_id Qwen/Qwen3-1.7B \\\n",
     "  --tokenizer_id Qwen/Qwen3-1.7B \\\n",
-    "  --output_dir ~/environment/ml/qwen \\\n",
+    "  --output_dir ~/ml/qwen \\\n",
     "  --bf16 True \\\n",
     "  --gradient_checkpointing True \\\n",
     "  --gradient_accumulation_steps 1 \\\n",
@@ -112,11 +112,11 @@
    "outputs": [],
    "source": [
     "!optimum-cli export neuron \\\n",
-    "  --model /home/ubuntu/environment/ml/qwen/merged_model \\\n",
+    "  --model ~/ml/qwen/merged_model \\\n",
     "  --task text-generation \\\n",
     "  --sequence_length 512 \\\n",
     "  --batch_size 1 \\\n",
-    "  /home/ubuntu/environment/ml/qwen/compiled_model"
+    "  ~/ml/qwen/compiled_model"
    ]
   },
   {
@@ -146,7 +146,7 @@
     "import os\n",
     "from vllm import LLM, SamplingParams\n",
     "llm = LLM(\n",
-    "    model=\"/home/ubuntu/environment/ml/qwen/compiled_model\", #local compiled model\n",
+    "    model=\"/home/ubuntu/ml/qwen/compiled_model\", #local compiled model\n",
     "    max_num_seqs=1,\n",
     "    max_model_len=2048,\n",
     "    device=\"neuron\",\n",

labs/FineTuning/HuggingFaceExample/02_inference/Inference-TinyLlama-1.1B.ipynb

@@ -200,7 +200,7 @@
     "s3_output_path=f\"{s3_orig_model_path}compiled_model/\"\n",
     "print(\"s3_output_path\",s3_output_path)\n",
     "training_job_name = utils.name_from_base(\"TGICompilation\")\n",
-    "print(\"training_job_name\")\n",
+    "print(\"training_job_name\",training_job_name)\n",
     "s3_model_path = f\"{s3_orig_model_path}merged_model/\"\n",
     "print(\"s3_model_path\",s3_model_path)\n",
     "\n",

labs/NxD/Lab_One_NxDI.ipynb

@@ -81,7 +81,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!hf download NousResearch/Llama-3.2-1B --local-dir /home/ubuntu/environment/models/llama/"
+    "!hf download NousResearch/Llama-3.2-1B --local-dir /home/ubuntu/build-on-trainium-workshop/labs/models/llama/"
    ]
   },
   {
@@ -101,7 +101,7 @@
    "outputs": [],
    "source": [
     "# the original checkpoint\n",
-    "model_path = '/home/ubuntu/environment/models/llama/'"
+    "model_path = '/home/ubuntu/build-on-trainium-workshop/labs/models/llama/'"
    ]
   },
   {
@@ -112,7 +112,7 @@
    "outputs": [],
    "source": [
     "# where your NxD trace will go\n",
-    "traced_model_path = '/home/ubuntu/environment/models/traced_llama'"
+    "traced_model_path = '/home/ubuntu/build-on-trainium-workshop/labs/models/traced_llama'"
    ]
   },
   {

labs/vLLM/Chess/Chess-Deployment.ipynb

@@ -109,7 +109,7 @@
     "import os\n",
     "\n",
     "# Option 1: Use your fine-tuned model (update path if you completed fine-tuning lab)\n",
-    "FINE_TUNED_MODEL_PATH = \"/home/ubuntu/environment/ml/qwen-chess/compiled_model\" # default path from previous lab\n",
+    "FINE_TUNED_MODEL_PATH = \"/home/ubuntu/ml/qwen-chess/compiled_model\" # default path from previous lab\n",
     "\n",
     "# Option 2: Use pre-compiled model from HF repository (default), this will be pulled down by optimum-neuron-vllm\n",
     "PRECOMPILED_MODEL_PATH = \"kunhunjon/ChessLM_Qwen3_Trainium_AWS_Format\"\n",
@@ -171,7 +171,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%cd /home/ubuntu/environment/vLLM/Chess/assets/\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/\n",
     "# Update vllm.sh with correct model path\n",
     "import os\n",
     "import re\n",
@@ -210,7 +210,7 @@
     "nohup bash vllm.sh > vllm-server.log 2>&1 &\n",
     "\n",
     "echo \"Server starting (PID: $!)\"\n",
-    "echo \"Check logs from /home/ubuntu/environment: tail -f vLLM/Chess/assets/vllm-server/vllm-server.log\"\n"
+    "echo \"Check logs from /home/ubuntu/build-on-trainium-workshop/labs: tail -f vLLM/Chess/assets/vllm-server/vllm-server.log\"\n"
    ]
   },
   {
@@ -221,7 +221,7 @@
     "\n",
     "You can monitor the server logs in the terminal:\n",
     "```bash\n",
-    "tail -f /home/ubuntu/environment/vLLM/Chess/assets/vllm-server/vllm-server.log\n",
+    "tail -f /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/vllm-server/vllm-server.log\n",
     "```\n",
     "\n",
     "Or check Neuron core usage:\n",
@@ -299,7 +299,7 @@
    "outputs": [],
    "source": [
     "# copy your env vars in\n",
-    "%cd /home/ubuntu/environment/vLLM/Chess/assets/\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/\n",
     "! cp env.example .env\n"
    ]
   },
@@ -313,7 +313,7 @@
     "%load_ext dotenv\n",
     "\n",
     "# 2. Load your .env file (adjust path if needed!)\n",
-    "%dotenv /home/ubuntu/environment/vLLM/Chess/assets/.env"
+    "%dotenv /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/.env"
    ]
   },
   {
@@ -347,7 +347,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%cd /home/ubuntu/environment/vLLM/Chess/\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/\n",
     "\n",
     "from assets.agents.vllm_agent import VLLMAgent\n",
     "from assets.agents.stockfish_agent import StockfishAgent\n",

labs/vLLM/Chess/Chess-Tournament.ipynb

@@ -73,7 +73,7 @@
    "outputs": [],
    "source": [
     "# copy your env vars in\n",
-    "%cd /home/ubuntu/environment/vLLM/Chess/assets/\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/\n",
     "! cp env.example .env\n"
    ]
   },
@@ -87,7 +87,7 @@
     "%load_ext dotenv\n",
     "\n",
     "# 2. Load your .env file (adjust path if needed!)\n",
-    "%dotenv /home/ubuntu/environment/vLLM/Chess/assets/.env"
+    "%dotenv /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/assets/.env"
    ]
   },
   {
@@ -123,7 +123,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%cd /home/ubuntu/environment/vLLM/Chess/\n",
+    "%cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess/\n",
     "\n",
     "# Play single game: vLLM vs Stockfish (skill 5)\n",
     "!python -m assets.run_game \\\n",

labs/vLLM/Chess/README.md

@@ -53,7 +53,7 @@ Learn how to:
 
 ```bash
 # Navigate to lab directory
-cd /home/ubuntu/environment/neuron-workshops/labs/vLLM/Chess
+cd /home/ubuntu/build-on-trainium-workshop/labs/vLLM/Chess
 
 #install stockfish
 sudo apt install stockfish