Initial functional tests

Co-authored-by: Trawinski, Dariusz <[email protected]>

tests/README.md

@@ -0,0 +1,18 @@
+# Running tests
+
+Installing and running tests
+```bash
+pip install -r requirements.txt
+pytest
+```
+
+Running tests with gpu
+```bash
+pytest --gpu
+```
+
+Run tests while caching downloaded models
+```bash
+pytest --model-cache ./cache
+```
+

tests/configs/dynamic.pbtxt

@@ -0,0 +1,20 @@
+input [
+  {
+    name: "data"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, -1, -1]
+  }
+]
+output [
+  {
+    name: "detection_out"
+    data_type: TYPE_FP32
+    dims: [ 1, 1, -1, 7]
+  }
+]
+parameters: {
+key: "RESHAPE_IO_LAYERS"
+value: {
+string_value:"yes"
+}
+}

tests/configs/dynamic_gpu.pbtxt

@@ -0,0 +1,28 @@
+input: [
+  {
+    name: "data"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, -1, -1]
+  }
+]
+output: [
+  {
+    name: "detection_out"
+    data_type: TYPE_FP32
+    dims: [ 1, 1, -1, 7]
+  }
+]
+parameters: [
+{
+    key: "RESHAPE_IO_LAYERS"
+    value: {
+      string_value:"yes"
+    }
+},
+{
+   key: "TARGET_DEVICE"
+   value: {
+     string_value: "GPU"
+   }
+}
+]

tests/configs/ir.pbtxt

@@ -0,0 +1,14 @@
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]

tests/configs/ir_gpu.pbtxt

@@ -0,0 +1,20 @@
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+},
+{
+   key: "TARGET_DEVICE"
+   value: {
+     string_value: "GPU"
+   }
+}
+]

tests/configs/onnx.pbtxt

@@ -0,0 +1,23 @@
+backend: "openvino"
+input: [
+  {
+    name: "gpu_0/data_0"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, 224, 224]
+  }
+]
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]
+default_model_filename: "model.onnx"

tests/configs/onnx_gpu.pbtxt

@@ -0,0 +1,30 @@
+backend: "openvino"
+input: [
+  {
+    name: "gpu_0/data_0"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, 224, 224]
+  }
+]
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+},
+{
+   key: "TARGET_DEVICE"
+   value: {
+     string_value: "GPU"
+   }
+}
+
+]
+default_model_filename: "model.onnx"

tests/configs/paddle.pbtxt

@@ -0,0 +1,31 @@
+backend: "openvino"
+input: [
+  {
+    name: "inputs"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, 224, 224]
+  }
+]
+output: [
+  {
+    name: "save_infer_model/scale_0.tmp_1"
+    data_type: TYPE_FP32
+    dims: [ -1, 1000]
+  }
+]
+
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]
+default_model_filename: "model.pdmodel"

tests/configs/paddle_gpu.pbtxt

@@ -0,0 +1,31 @@
+backend: "openvino"
+input: [
+  {
+    name: "inputs"
+    data_type: TYPE_FP32
+    dims: [ -1, 3, 224, 224]
+  }
+]
+output: [
+  {
+    name: "save_infer_model/scale_0.tmp_1"
+    data_type: TYPE_FP32
+    dims: [ -1, 1000]
+  }
+]
+
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]
+default_model_filename: "model.pdmodel"

tests/configs/saved_model.pbtxt

@@ -0,0 +1,30 @@
+backend: "openvino"
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]
+input [
+  {
+    name: "input_1"
+    data_type: TYPE_FP32
+    dims: [-1, 224, 224, 3 ]
+  }
+]
+output [
+  {
+    name: "activation_49"
+    data_type: TYPE_FP32
+    dims: [-1, 1001]
+  }
+]
+default_model_filename: "model.saved_model"

tests/configs/tflite.pbtxt

@@ -0,0 +1,16 @@
+backend: "openvino"
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+}
+]
+default_model_filename: "model.tflite"

tests/configs/tflite_gpu.pbtxt

@@ -0,0 +1,22 @@
+backend: "openvino"
+parameters: [
+{
+   key: "NUM_STREAMS"
+   value: {
+     string_value: "1"
+   }
+},
+{
+   key: "PERFORMANCE_HINT"
+   value: {
+     string_value: "LATENCY"
+   }
+},
+{
+   key: "TARGET_DEVICE"
+   value: {
+     string_value: "GPU"
+   }
+}
+]
+default_model_filename: "model.tflite"

tests/conftest.py

@@ -0,0 +1,29 @@
+import shutil
+import subprocess
+import sys
+
+import pytest
+
+
+def pytest_addoption(parser):
+    parser.addoption("--gpu", action="store_true")
+    parser.addoption("--model-cache", action="store")
+    parser.addoption("--image", action="store")
+
+
+def pytest_configure(config):
+    config.addinivalue_line(
+        "markers",
+        "gpu: run tests on GPU device",
+    )
+
+
+def pytest_collection_modifyitems(config, items):
+    for item in items:
+        if "gpu" in item.keywords:
+            if not config.getoption("--gpu"):
+                item.add_marker(
+                    pytest.mark.skip(
+                        "Test requires --gpu flag to be set and Intel GPU device on the host machine"
+                    )
+                )

tests/functional/common.py

@@ -0,0 +1,26 @@
+import os
+
+import grpc
+import numpy as np
+import tritonclient.grpc as grpcclient
+from tritonclient.grpc import service_pb2_grpc
+from tritonclient.utils import np_to_triton_dtype
+
+
+def prepare_inputs(input_name, shape, dtype):
+    inputs = []
+    inputs.append(grpcclient.InferInput(input_name, shape, np_to_triton_dtype(dtype)))
+
+    inputs[0].set_data_from_numpy(np.ones(shape, dtype=dtype))
+    return inputs
+
+
+def prepare_grpc_stub(port):
+    channel = grpc.insecure_channel(f"localhost:{port}")
+    return service_pb2_grpc.GRPCInferenceServiceStub(channel)
+
+
+def prepare_triton_client(port):
+    return grpcclient.InferenceServerClient(
+        url=f"localhost:{port}", verbose=os.environ.get("LOG_LEVEL") == "DEBUG"
+    )