import test.test_utils.ec2 as ec2_utils from test import test_utils from test.test_utils import CONTAINER_TESTS_PREFIX, get_framework_and_version_from_tag from test.test_utils.ec2 import get_ec2_instance_type, execute_ec2_inference_test, get_ec2_accelerator_type from test.dlc_tests.conftest import LOGGER SQUEEZENET_MODEL = "squeezenet" BERT_MODEL = "bert_sst" RESNET_EIA_MODEL = "resnet-152-eia" MX_EC2_GPU_INSTANCE_TYPE = get_ec2_instance_type(default="g3.8xlarge", processor="gpu") MX_EC2_CPU_INSTANCE_TYPE = get_ec2_instance_type(default="c5.4xlarge", processor="cpu") MX_EC2_EIA_ACCELERATOR_TYPE = get_ec2_accelerator_type(default="eia1.large", processor="eia") MX_EC2_GPU_EIA_INSTANCE_TYPE = get_ec2_instance_type( default="g3.8xlarge", processor="gpu", filter_function=ec2_utils.filter_not_heavy_instance_types, ) MX_EC2_SINGLE_GPU_INSTANCE_TYPE = get_ec2_instance_type( default="p3.2xlarge", processor="gpu", filter_function=ec2_utils.filter_only_single_gpu, ) MX_EC2_NEURON_INSTANCE_TYPE = get_ec2_instance_type(default="inf1.xlarge", processor="neuron") MX_TELEMETRY_CMD = os.path.join(CONTAINER_TESTS_PREFIX, "test_mx_dlc_telemetry_test")
import os import pytest from test import test_utils from test.test_utils import CONTAINER_TESTS_PREFIX, get_framework_and_version_from_tag from test.test_utils.ec2 import get_ec2_instance_type, execute_ec2_inference_test, get_ec2_accelerator_type from test.dlc_tests.conftest import LOGGER PT_EC2_GPU_INSTANCE_TYPE = get_ec2_instance_type(default="g3.8xlarge", processor="gpu") PT_EC2_CPU_INSTANCE_TYPE = get_ec2_instance_type(default="c5.9xlarge", processor="cpu") PT_EC2_EIA_ACCELERATOR_TYPE = get_ec2_accelerator_type(default="eia1.large", processor="eia") PT_TELEMETRY_CMD = os.path.join(CONTAINER_TESTS_PREFIX, "pytorch_tests", "test_pt_dlc_telemetry_test") PT_EC2_NEURON_ACCELERATOR_TYPE = get_ec2_accelerator_type( default="inf1.xlarge", processor="neuron") @pytest.mark.model("resnet") @pytest.mark.parametrize("ec2_instance_ami", [test_utils.NEURON_AL2_DLAMI], indirect=True) @pytest.mark.parametrize("ec2_instance_type", PT_EC2_NEURON_ACCELERATOR_TYPE, indirect=True) def test_ec2_pytorch_inference_gpu(pytorch_inference, ec2_connection, region, gpu_only): ec2_pytorch_inference(pytorch_inference, "neuron", ec2_connection, region)