64 lines
2.2 KiB
Python
64 lines
2.2 KiB
Python
|
"""
|
||
|
|
Script to initialize an Azure Machine Learning compute cluster (aml)
|
||
|
|
"""
|
||
|
|
from azure.ai.ml.entities import AmlCompute
|
||
|
|
|
||
|
|
from initialize_constants import AML_COMPUTE_NAME, MAX_NODES, MIN_NODES, VM_SIZE
|
||
|
|
from ml_client import create_or_load_ml_client
|
||
|
|
|
||
|
|
|
||
|
|
def create_or_load_aml(
|
||
|
|
cpu_compute_target=AML_COMPUTE_NAME,
|
||
|
|
vm_size=VM_SIZE,
|
||
|
|
min_nodes=MIN_NODES,
|
||
|
|
max_nodes=MAX_NODES,
|
||
|
|
):
|
||
|
|
"""Create or load an Azure Machine Learning compute cluster (aml) in a
|
||
|
|
given Workspace.
|
||
|
|
Args:
|
||
|
|
cpu_compute_target: Name of the compute resource
|
||
|
|
vm_size: Virtual machine size, VM_SIZE is used as default,
|
||
|
|
for example STANDARD_D2_V2. Set to STANDARD_NC6 to get a GPU
|
||
|
|
min_nodes: Minimal number of nodes, MIN_NODES is used as default.
|
||
|
|
max_nodes: Minimal number of nodes, MIN_NODES is used as default.
|
||
|
|
|
||
|
|
Returns:
|
||
|
|
An aml and set quick load.
|
||
|
|
"""
|
||
|
|
# Create or Load a Workspace
|
||
|
|
ml_client = create_or_load_ml_client()
|
||
|
|
try:
|
||
|
|
# let's see if the compute target already exists
|
||
|
|
cpu_cluster = ml_client.compute.get(cpu_compute_target)
|
||
|
|
print(
|
||
|
|
f"You already have a cluster named {cpu_compute_target},",
|
||
|
|
"we'll reuse it.",
|
||
|
|
)
|
||
|
|
except Exception:
|
||
|
|
print("Creating a new cpu compute target...")
|
||
|
|
cpu_cluster = AmlCompute(
|
||
|
|
name=cpu_compute_target,
|
||
|
|
# Azure ML Compute is the on-demand VM service
|
||
|
|
type="amlcompute",
|
||
|
|
# VM Family
|
||
|
|
size=vm_size,
|
||
|
|
# Minimum running nodes when there is no job running
|
||
|
|
min_instances=min_nodes,
|
||
|
|
# Nodes in cluster
|
||
|
|
max_instances=max_nodes,
|
||
|
|
# How many seconds will the node running after the job termination
|
||
|
|
idle_time_before_scale_down=180,
|
||
|
|
# Dedicated or LowPriority.
|
||
|
|
# The latter is cheaper but there is a chance of job termination
|
||
|
|
tier="Dedicated",
|
||
|
|
)
|
||
|
|
|
||
|
|
# Now, we pass the object to MLClient's create_or_update method
|
||
|
|
cpu_cluster = ml_client.compute.begin_create_or_update(cpu_cluster)
|
||
|
|
|
||
|
|
return cpu_cluster
|
||
|
|
|
||
|
|
|
||
|
|
if __name__ == "__main__":
|
||
|
|
create_or_load_aml()
|