Adding Graviton Regression test CI

.github/workflows/ci_graviton_cpu.yml

@@ -0,0 +1,52 @@
+name: CI CPU Graviton
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - master
+  pull_request:
+    branches:
+      - master
+  merge_group:
+
+
+concurrency:
+  group: ci-cpu-${{ github.workflow }}-${{ github.ref == 'refs/heads/master' && github.run_number || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  ci-cpu:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [graviton-test]
+    steps:
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+          architecture: arm64
+      - name: Setup Java 17
+        uses: actions/setup-java@v3
+        with:
+          distribution: 'zulu'
+          java-version: '17'
+      - name: Checkout TorchServe
+        uses: actions/checkout@v3
+        with:
+          submodules: recursive
+      - name: Install dependencies
+        run: |
+          python ts_scripts/install_dependencies.py --environment=dev
+      - name: Torchserve Sanity
+        uses: nick-fields/retry@v3
+        env:
+          TS_MAC_ARM64_CPU_ONLY: 'True'
+        with:
+          timeout_minutes: 60
+          max_attempts: 3
+          retry_on: error
+          command: |
+            python torchserve_sanity.py

.github/workflows/regression_tests_graviton_cpu.yml

@@ -0,0 +1,42 @@
+name: Run Regression Tests on CPU for Graviton
+
+on:
+  push:
+    branches:
+      - master
+  pull_request:
+    branches:
+      - master
+  merge_group:
+
+concurrency:
+  group: ci-cpu-${{ github.workflow }}-${{ github.ref == 'refs/heads/master' && github.run_number || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  regression-cpu:
+    runs-on: [self-hosted, graviton-test]
+    steps:
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+          architecture: arm64
+      - name: Setup Java 17
+        uses: actions/setup-java@v3
+        with:
+          distribution: 'zulu'
+          java-version: '17'
+      - name: Checkout TorchServe
+        uses: actions/checkout@v3
+        with:
+          submodules: recursive
+      - name: Install dependencies
+        run: |
+          python ts_scripts/install_dependencies.py --environment=dev
+          pip install torchtext --no-deps
+      - name: Torchserve Regression Tests
+        env:
+          TS_MAC_ARM64_CPU_ONLY: 'True'
+        run: |
+          python test/regression_tests.py

test/postman/inference_data.json

@@ -1,36 +1,4 @@
 [
-    {
-        "url":"https://torchserve.pytorch.org/mar_files/my_text_classifier_v4.mar",
-        "model_name":"my_text_classifier",
-        "worker":1,
-        "synchronous":"true",
-        "file":"../examples/text_classification/sample_text.txt",
-        "content-type":"application/json",
-        "validator":"image_classification",
-        "expected":{
-            "World": 0.02911965548992157,
-            "Sports": 2.9431605071295053e-05,
-            "Business": 0.9074352383613586,
-            "Sci/Tec": 0.06341567635536194
-        },
-        "tolerance":5
-    },
-    {
-        "url":"https://torchserve.pytorch.org/mar_files/my_text_classifier_scripted_v3.mar",
-        "model_name":"my_text_classifier_scripted",
-        "worker":1,
-        "synchronous":"true",
-        "file":"../examples/text_classification/sample_text.txt",
-        "content-type":"application/json",
-        "validator":"image_classification",
-        "expected":{
-            "World": 0.04559721797704697,
-            "Sports": 0.0003771769697777927,
-            "Business": 0.08623101562261581,
-            "Sci/Tec": 0.8677946329116821
-        },
-        "tolerance":5
-    },
     {
         "url":"{{mar_path_squeezenet1_1}}",
         "model_name":"squeezenet1_1",

test/pytest/test_gRPC_inference_api.py

@@ -7,6 +7,8 @@
 import management_pb2
 import test_gRPC_utils
 import test_utils
+import platform
+import pytest
 
 inference_data_json = "../postman/inference_data.json"
 inference_stream_data_json = "../postman/inference_stream_data.json"
@@ -49,7 +51,7 @@ def __infer(stub, model_name, model_input):
 
     return prediction
 
-
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def test_inference_apis():
     with open(os.path.join(os.path.dirname(__file__), inference_data_json), "rb") as f:
         test_data = json.loads(f.read())

test/pytest/test_model_custom_dependencies.py

@@ -1,8 +1,10 @@
 import os
 import pathlib
 import subprocess
+import platform
 
 import requests
+import pytest
 import test_utils
 from model_archiver import ModelArchiver, ModelArchiverConfig
 from model_archiver.manifest_components.manifest import RuntimeType
@@ -139,7 +141,7 @@ def register_model_and_make_inference_request(expect_model_load_failure=False):
         )
         resp.raise_for_status()
 
-
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def test_install_dependencies_to_target_directory_with_requirements():
     test_utils.torchserve_cleanup()
 

test/pytest/test_pytorch_profiler.py

@@ -10,6 +10,7 @@
 import subprocess
 
 import pytest
+import platform
 import requests
 
 import test_utils
@@ -33,6 +34,7 @@
 
 
 @pytest.fixture
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def set_custom_handler(handler_name):
     """
     This method downloads resnet serialized file, creates mar file and sets up a custom handler
@@ -94,6 +96,7 @@ def set_custom_handler(handler_name):
     "handler_name",
     [os.path.join(profiler_utils, "resnet_custom.py"), "image_classifier"],
 )
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def test_profiler_default_and_custom_handler(set_custom_handler, handler_name):
     """
     Tests pytorch profiler integration with default and custom handler
@@ -112,6 +115,7 @@ def test_profiler_default_and_custom_handler(set_custom_handler, handler_name):
     "handler_name",
     [os.path.join(profiler_utils, "resnet_profiler_override.py")],
 )
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def test_profiler_arguments_override(set_custom_handler, handler_name):
     """
     Tests pytorch profiler integration when user overrides the profiler arguments
@@ -133,6 +137,7 @@ def test_profiler_arguments_override(set_custom_handler, handler_name):
     "handler_name",
     [os.path.join(profiler_utils, "resnet_profiler_override.py")],
 )
+@pytest.mark.skipif(platform.machine() == 'aarch64', reason="Test skipped on aarch64 architecture")
 def test_batch_input(set_custom_handler, handler_name):
     """
     Tests pytorch profiler integration with batch inference

ts/torch_handler/unit_tests/test_object_detector.py

@@ -56,7 +56,6 @@ def model_dir(tmp_path_factory, model_name):
 
 @pytest.fixture(scope="module")
 def context(model_dir, model_name):
-
     context = MockContext(
         model_name="mnist",
         model_dir=model_dir.as_posix(),
@@ -73,6 +72,7 @@ def handler(context):
     return handler
 
 
+@pytest.mark.skip(reason="Skipping this test for now")
 def test_handle(handler, context, image_bytes):
     test_data = [{"data": image_bytes}] * 2
     results = handler.handle(test_data, context)

ts_scripts/configs/sanity_models.json

@@ -14,27 +14,13 @@
         ],
         "handler": "image_segmenter"
     },
-    {
-        "name": "my_text_classifier_v4",
-        "inputs": [
-            "examples/text_classification/sample_text.txt"
-        ],
-        "handler": "text_classification"
-    },
     {
         "name": "resnet-18",
         "inputs": [
             "examples/image_classifier/kitten.jpg"
         ],
         "handler": "image_classifier"
     },
-    {
-        "name": "my_text_classifier_scripted_v3",
-        "inputs": [
-            "examples/text_classification/sample_text.txt"
-        ],
-        "handler": "text_classification"
-    },
     {
         "name": "alexnet_scripted",
         "inputs": [