Spleen Segmentation Example - NVIDIA/clara-platform-python-client GitHub Wiki

This is a full example demonstrating running a spleen segmentation pipeline. This includes creating the pipeline, launching the job, and interacting with payloads.

>>> from nvidia_clara.jobs_client import JobsClient
>>> from nvidia_clara.pipelines_client import PipelinesClient
>>> from nvidia_clara.payloads_client import PayloadsClient
>>> import nvidia_clara.pipeline_types as pipeline_types
>>> import os
>>> from pathlib import Path

# Clients creation
>>> clara_ip_address = "10.0.0.1"
>>> clara_port = "30407"

>>> jobs_client = JobsClient(target=clara_ip_address, port=clara_port)
>>> payloads_client = PayloadsClient(target=clara_ip_address, port=clara_port)
>>> pipelines_client = PipelinesClient(target=clara_ip_address, port=clara_port)

# Create list of pipeline_types.PipelineDefinition with local path to pipeline .yaml
>>> file_path = "./spleen_pipeline.yaml"
>>> definitions = [pipeline_types.PipelineDefinition(name=file_path, content=Path(file_path).read_text())]

# Create Pipeline with definition list created
>>> pipeline_id = pipelines_client.create_pipeline(definition=definitions)

# Create Job with newly created Pipeline
>>> job_info = jobs_client.create_job(job_name="spleenjob", pipeline_id=pipeline_types.PipelineId(pipeline_id.value))
>>> job_id = job_info.job_id
>>> payload_id = job_info.payload_id

# Local path to directory of files to upload to the job's payload on the Server
>>> input_path = "./app_spleen-input_v1/dcm"

# Go through files in directory and upload them as input to the job using the payload identifier
>>> for file in os.listdir(input_path):
...     file_path = input_path + "/" + str(file)
...     with open(file_path, 'rb') as fp:
...         payloads_client.upload(payload_id=payload_id, blob_name=file, file_object=fp)

# Get a List of the jobs
>>> job_list = jobs_client.list_jobs()

# Start Job
>>> job_token = jobs_client.start_job(job_id=job_id)

# Loop until job completes
>>> job_status = jobs_client.get_status(job_id=job_id)
>>> while job_status.job_state != 3:
...     job_status = jobs_client.get_status(job_id=job_id)

# Get Payload Details - Used to get list of payload files
>>> payload_details = payloads_client.get_details(payload_id=payload_id)

# Download files from payload if pertaining to output payload directory (ex. "/operators)
>>> for file in payload_details.file_details:
...
...     # Get file path on Server (ex. /operators/dicom-reader/example_file.raw")
...     file_name = file.name
...
...     # Split file path name (ex. ['','operators','dicom-reader','example_file.raw'])
...     name = file_name.split('/')
...
...     # Check if file pertains to output directory (ex. "/operators")
...     if name[1] == 'operators':
...
...         # Download file to a local results directory to a file with same name on server (ex. example_file.raw)
...         with open("./results/"+name[-1], 'wb+') as wb:
...             payloads_client.download_from(payload_id=payload_id, blob_name="."+file_name, dest_obj=wb)


# Gets list of operator logs from job
>>> jobs_logs = jobs_client.job_logs(job_id=job_id, operator_name="dicom-reader")