def test_sdk_e2e():
container = V1Container(
name="xgboost",
image="docker.io/merlintang/xgboost-dist-iris:1.1",
args=[
"--job_type=Train",
"--xgboost_parameter=objective:multi:softprob,num_class:3",
"--n_estimators=10", "--learning_rate=0.1",
"--model_path=/tmp/xgboost-model", "--model_storage_type=local"
],
)
master = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(spec=V1PodSpec(containers=[container])))
worker = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(spec=V1PodSpec(containers=[container])))
xgboostjob = KubeflowOrgV1XGBoostJob(
api_version="kubeflow.org/v1",
kind="XGBoostJob",
metadata=V1ObjectMeta(name="xgboostjob-iris-ci-test",
namespace=SDK_TEST_NAMESPACE),
spec=KubeflowOrgV1XGBoostJobSpec(run_policy=V1RunPolicy(
clean_pod_policy="None", ),
xgb_replica_specs={
"Master": master,
"Worker": worker
}))
XGBOOST_CLIENT.create(xgboostjob)
XGBOOST_CLIENT.wait_for_job("xgboostjob-iris-ci-test",
namespace=SDK_TEST_NAMESPACE)
if not XGBOOST_CLIENT.is_job_succeeded("xgboostjob-iris-ci-test",
namespace=SDK_TEST_NAMESPACE):
raise RuntimeError("The XGBoostJob is not succeeded.")
XGBOOST_CLIENT.get_logs("xgboostjob-iris-ci-test",
namespace=SDK_TEST_NAMESPACE)
XGBOOST_CLIENT.delete("xgboostjob-iris-ci-test",
namespace=SDK_TEST_NAMESPACE)
def test_sdk_e2e():
container = V1Container(
name="pytorch",
image="gcr.io/kubeflow-ci/pytorch-dist-mnist-test:v1.0",
args=["--backend", "gloo"],
)
master = V1ReplicaSpec(
replicas=1,
restart_policy="OnFailure",
template=V1PodTemplateSpec(spec=V1PodSpec(containers=[container])))
worker = V1ReplicaSpec(
replicas=1,
restart_policy="OnFailure",
template=V1PodTemplateSpec(spec=V1PodSpec(containers=[container])))
pytorchjob = KubeflowOrgV1PyTorchJob(
api_version="kubeflow.org/v1",
kind="PyTorchJob",
metadata=V1ObjectMeta(name="pytorchjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE),
spec=KubeflowOrgV1PyTorchJobSpec(run_policy=V1RunPolicy(
clean_pod_policy="None", ),
pytorch_replica_specs={
"Master": master,
"Worker": worker
}))
PYTORCH_CLIENT.create(pytorchjob)
PYTORCH_CLIENT.wait_for_job("pytorchjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE)
if not PYTORCH_CLIENT.is_job_succeeded("pytorchjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE):
raise RuntimeError("The PyTorchJob is not succeeded.")
PYTORCH_CLIENT.get_logs("pytorchjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE)
PYTORCH_CLIENT.delete("pytorchjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE)
def test_sdk_e2e():
container = V1Container(
name="tensorflow",
image="gcr.io/kubeflow-ci/tf-mnist-with-summaries:1.0",
command=[
"python",
"/var/tf_mnist/mnist_with_summaries.py",
"--log_dir=/train/logs", "--learning_rate=0.01",
"--batch_size=150"
]
)
worker = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(
spec=V1PodSpec(
containers=[container]
)
)
)
tfjob = V1TFJob(
api_version="kubeflow.org/v1",
kind="TFJob",
metadata=V1ObjectMeta(name="mnist-ci-test", namespace=SDK_TEST_NAMESPACE),
spec=V1TFJobSpec(
run_policy=V1RunPolicy(
clean_pod_policy="None",
),
tf_replica_specs={"Worker": worker}
)
)
TFJOB_CLIENT.create(tfjob, namespace=SDK_TEST_NAMESPACE)
TFJOB_CLIENT.wait_for_job("mnist-ci-test", namespace=SDK_TEST_NAMESPACE)
if not TFJOB_CLIENT.is_job_succeeded("mnist-ci-test", namespace=SDK_TEST_NAMESPACE):
raise RuntimeError("The TFJob is not succeeded.")
TFJOB_CLIENT.get_logs("mnist-ci-test", master=False, namespace=SDK_TEST_NAMESPACE)
TFJOB_CLIENT.delete("mnist-ci-test", namespace=SDK_TEST_NAMESPACE)
def test_sdk_e2e():
master_container = V1Container(
name="mpi",
image="horovod/horovod:0.20.0-tf2.3.0-torch1.6.0-mxnet1.5.0-py3.7-cpu",
command=["mpirun"],
args=[
"-np",
"1",
"--allow-run-as-root",
"-bind-to",
"none",
"-map-by",
"slot",
"-x",
"LD_LIBRARY_PATH",
"-x",
"PATH",
"-mca",
"pml",
"ob1",
"-mca",
"btl",
"^openib",
#"python", "/examples/tensorflow2_mnist.py"]
"python",
"/examples/pytorch_mnist.py",
"--epochs",
"1"
])
worker_container = V1Container(
name="mpi",
image="horovod/horovod:0.20.0-tf2.3.0-torch1.6.0-mxnet1.5.0-py3.7-cpu",
)
master = V1ReplicaSpec(replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(spec=V1PodSpec(
containers=[master_container])))
worker = V1ReplicaSpec(replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(spec=V1PodSpec(
containers=[worker_container])))
mpijob = KubeflowOrgV1MPIJob(
api_version="kubeflow.org/v1",
kind="MPIJob",
metadata=V1ObjectMeta(name="mpijob-mxnet-ci-test",
namespace=SDK_TEST_NAMESPACE),
spec=KubeflowOrgV1MPIJobSpec(slots_per_worker=1,
run_policy=V1RunPolicy(
clean_pod_policy="None", ),
mpi_replica_specs={
"Launcher": master,
"Worker": worker
}))
MPI_CLIENT.create(mpijob)
MPI_CLIENT.wait_for_job("mpijob-mxnet-ci-test",
namespace=SDK_TEST_NAMESPACE)
if not MPI_CLIENT.is_job_succeeded("mpijob-mxnet-ci-test",
namespace=SDK_TEST_NAMESPACE):
raise RuntimeError("The MPIJob is not succeeded.")
MPI_CLIENT.get_logs("mpijob-mxnet-ci-test", namespace=SDK_TEST_NAMESPACE)
MPI_CLIENT.delete("mpijob-mxnet-ci-test", namespace=SDK_TEST_NAMESPACE)
def test_sdk_e2e():
worker_container = V1Container(
name="mxnet",
image="docker.io/johnugeorge/mxnet:1.9.1_cpu_py3",
command=["/usr/local/bin/python3"],
args=["incubator-mxnet/example/image-classification/train_mnist.py",
"--num-epochs", "5",
"--num-examples","1000",
"--kv-store", "dist_sync"],
ports=[V1ContainerPort(container_port=9991, name="mxjob-port")]
)
server_container = V1Container(
name="mxnet",
image="docker.io/johnugeorge/mxnet:1.9.1_cpu_py3",
ports=[V1ContainerPort(container_port=9991, name="mxjob-port")]
)
scheduler_container = V1Container(
name="mxnet",
image="docker.io/johnugeorge/mxnet:1.9.1_cpu_py3",
ports=[V1ContainerPort(container_port=9991, name="mxjob-port")]
)
worker = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(
spec=V1PodSpec(
containers=[worker_container]
)
)
)
server = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(
spec=V1PodSpec(
containers=[server_container]
)
)
)
scheduler = V1ReplicaSpec(
replicas=1,
restart_policy="Never",
template=V1PodTemplateSpec(
spec=V1PodSpec(
containers=[scheduler_container]
)
)
)
mxjob = KubeflowOrgV1MXJob(
api_version="kubeflow.org/v1",
kind="MXJob",
metadata=V1ObjectMeta(name="mxjob-mnist-ci-test", namespace=SDK_TEST_NAMESPACE),
spec=KubeflowOrgV1MXJobSpec(
job_mode="MXTrain",
run_policy=V1RunPolicy(
clean_pod_policy="None",
),
mx_replica_specs={"Scheduler": scheduler,
"Server": server,
"Worker": worker}
)
)
MX_CLIENT.create(mxjob)
MX_CLIENT.wait_for_job("mxjob-mnist-ci-test", namespace=SDK_TEST_NAMESPACE)
if not MX_CLIENT.is_job_succeeded("mxjob-mnist-ci-test",
namespace=SDK_TEST_NAMESPACE):
raise RuntimeError("The MXJob is not succeeded.")
MX_CLIENT.get_logs("mxjob-mnist-ci-test", namespace=SDK_TEST_NAMESPACE, master=False)
MX_CLIENT.delete("mxjob-mnist-ci-test", namespace=SDK_TEST_NAMESPACE)