Training¶

Helper functions for training.

Method	Description
experiments_create	Create experiment
experiments_delete	Delete experiment
experiments_get	Get experiment
experiments_list	List experiments
experiments_update	Update experiment
experiment_runs_create	Create experiment run
experiment_runs_delete	Delete experiment run
experiment_runs_get	Get experiment run
experiment_runs_list	List experiment runs
experiment_runs_update	Update experiment run
experiment_environment_variables_create	Create experiment environment variable
experiment_environment_variables_delete	Delete experiment environment variable
experiment_environment_variables_get	Get experiment environment variable
experiment_environment_variables_list	List experiment environment variables
experiment_environment_variables_update	Update experiment environment variable

experiments_create¶

ExperimentList experiments_create(project_name, data)

Create a training experiment

Description¶

Helper function to create a training experiment

Required Parameters¶

name: The name of the experiment to create
environment: The name of the runtime environment to use

Optional Parameters¶

default_bucket: In which bucket to store the in- and output artifacts of the training experiment
description: Description of the experiment
instance_type: The instance type to allocate for this experiment
request_retention_time: The request retention time for this experiment
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label

Request Examples¶

{
  "name": "my-experiment",
  "environment": "python3-8"
}

{
  "name": "my-experiment",
  "description": "A training experiment",
  "environment": "python3-8",
  "instance_type": "4096mb",
  "default_bucket": "default",
  "labels": {
    "type": "pytorch"
  }
}

Response Structure¶

Details of the created experiment

id: Unique identifier for the experiment (UUID)
name: Experiment name
environment: Environment of the experiment
environment_display_name: Human readable name of the environment
status: The status of the experiment
active_revision: UUID of the active revision of the experiment
latest_revision: UUID of the latest revision of the experiment
instance_type: The instance type for the experiment
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label
creation_date: The date when the experiment was created
last_updated: The date when the experiment was last updated

Response Examples¶

{
  "id": "4ae7d14b-4803-4e16-b96d-3b18caa4b605",
  "name": "training-experiment",
  "environment": "python3-8",
  "environment_display_name": "Python 3.8",
  "status": "available",
  "active_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "latest_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "instance_type": "4069mb",
  "labels": {
    "type": "pytorch"
  },
  "creation_date": "2020-05-12T16:23:15.456812Z",
  "last_updated": "2020-05-12T16:23:15.456812Z"
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Create experiment
api_response = training.experiments_create(
    project_name=project_name,
    data=ubiops.ExperimentCreate(
        name=experiment_name,
        instance_type='4069mb',
        description='A training experiment',
        environment='python3-8',
        default_bucket='default',
        labels={"type": "pytorch"}
    )
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Create experiment
api_response = training.experiments_create(
    project_name=project_name,
    data=ubiops.ExperimentCreate(
        name=experiment_name,
        instance_type='4069mb',
        description='A training experiment',
        environment='python3-8',
        default_bucket='default',
        labels={"type": "pytorch"}
    )
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
data	ExperimentCreate

Return type¶

ExperimentList

Authorization¶

API token

[Back to top]

experiments_delete¶

experiments_delete(project_name, experiment_name)

Delete a training experiment

Description¶

Helper function to delete a training experiment

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Delete experiment
training.experiments_delete(project_name=project_name, experiment_name=experiment_name)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Delete experiment
training.experiments_delete(project_name=project_name, experiment_name=experiment_name)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str

Return type¶

void (empty response body)

Authorization¶

API token

[Back to top]

experiments_get¶

ExperimentDetail experiments_get(project_name, experiment_name)

Get the details of a training experiment

Description¶

Helper function to get the details of a training experiment

Response Structure¶

Get the details of an experiment

id: Unique identifier for the experiment (UUID)
name: Experiment name
description: Description of the experiment
environment: Environment of the experiment
environment_display_name: Human readable name of the environment
status: The status of the experiment
active_revision: UUID of the active revision of the experiment
latest_revision: UUID of the latest revision of the experiment
instance_type: The instance type for the experiment
maximum_instances: Upper bound of number of experiment pods running in parallel
minimum_instances: Lower bound of number of experiment pods running in parallel
maximum_idle_time: Maximum time in seconds a experiment stays idle before it is stopped
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label
creation_date: The date when the experiment was created
last_updated: The date when the experiment was last updated
monitoring: Name of a notification group which contains contacts to send notifications when runs for the experiment fail and recover
default_notification_group: Name of a notification group which contains contacts to send notifications when runs for the experiment are completed
request_retention_time: Number of seconds to store runs of the experiment
request_retention_mode: Mode of retention for runs to the experiment. It can be one of the following:
- none - the runs will not be stored
- metadata - only the metadata of the runs will be stored
- full - both the metadata and input/output of the runs will be stored
maximum_queue_size_express: Maximum number of queued express runs for all instances of this experiment
maximum_queue_size_batch: Maximum number of queued batch runs for all instances of this experiment
has_request_method: Whether the latest revision of the experiment has a 'request' method
has_requests_method: Whether the latest revision of the experiment has a 'requests' method
static_ip: A boolean indicating whether the experiment should get a static IP
restart_request_interruption: A boolean indicating whether the runs should be restarted in case of an interruption
default_bucket: Indicates in which bucket input and output artifacts of the training job will be stored

Response Examples¶

{
  "active_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "creation_date": "2021-06-15T10:12:11.554682Z",
  "default_bucket": "default",
  "default_notification_group": None,
  "description": "An experiment with Python 3.7",
  "environment": "python3-7",
  "environment_display_name": "Python 3.7",
  "has_request_method": True,
  "has_requests_method": False,
  "id": "49d7d195-2296-43c0-8eab-84b0acd6e5bc",
  "instance_type": "2048mb",
  "labels": {},
  "last_updated": 2021-06-15T10:12:11.554682Z,
  "latest_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "maximum_idle_time": 300,
  "maximum_instances": 10,
  "maximum_queue_size_batch": 100000,
  "maximum_queue_size_express": 100,
  "minimum_instances": 0,
  "monitoring": None,
  "name": "training-experiment",
  "request_retention_mode": "full",
  "request_retention_time": 31536000,
  "restart_request_interruption": False,
  "static_ip": False,
  "status": "available"
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = "experiment_name_example" # str

# Get experiment details
api_response = training.experiments_get(project_name=project_name, experiment_name=experiment_name)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Get experiment details
api_response = training.experiments_list(project_name=project_name, experiment_name=experiment_name)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str

Return type¶

[ExperimentDetail]

Authorization¶

API token

[Back to top]

experiments_list¶

list[ExperimentList] experiments_list(project_name)

List training experiments

Description¶

Helper function to list training experiments

Response Structure¶

A list of details of the experiments

id: Unique identifier for the experiment (UUID)
name: Experiment name
environment: Environment of the experiment
environment_display_name: Human readable name of the environment
status: The status of the experiment
active_revision: UUID of the active revision of the experiment
latest_revision: UUID of the latest revision of the experiment
instance_type: The instance type for the experiment
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label
creation_date: The date when the experiment was created
last_updated: The date when the experiment was last updated

Response Examples¶

[
  {
    "id": "4ae7d14b-4803-4e16-b96d-3b18caa4b605",
    "name": "experiment_name_example_1",
    "environment": "python3-8",
    "environment_display_name": "Python 3.8",
    "status": "available",
    "active_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
    "latest_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
    "instance_type": "4069mb",
    "labels": {
      "type": "pytorch"
    },
    "creation_date": "2020-05-12T16:23:15.456812Z",
    "last_updated": "2020-05-12T16:23:15.456812Z"
  },
  {
    "id": "536b4b33-b1db-4446-b07c-986806478654",
    "name": "experiment_name_example_2",
    "environment": "python3-7",
    "environment_display_name": "Python 3.7",
    "status": "available",
    "active_revision": "cd2244a7-5953-4d31-add8-e88cf5f4bb30",
    "latest_revision": "cd2244a7-5953-4d31-add8-e88cf5f4bb30",
    "instance_type": "16384mb",
    "labels": {
      "type": "tensorflow"
    },
    "creation_date": "2021-06-15T10:12:11.554682Z",
    "last_updated": "2021-06-15T10:12:11.554682Z"
  }
]

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str

# List experiments
api_response = training.experiments_list(project_name=project_name)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str

# List experiments
api_response = training.experiments_list(project_name=project_name)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str

Return type¶

list[ExperimentList]

Authorization¶

API token

[Back to top]

experiments_update¶

ExperimentDetail experiments_update(project_name, experiment_name, data)

Update a training experiment

Description¶

Helper function to update a training experiment

Optional Parameters¶

name: New name for the experiment
environment: New environment for the experiment. It can be either a base or a custom environment.
default_bucket: New default bucket for the experiment
description: New description for the experiment
instance_type: New instance type for the experiment
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label. The new labels will replace the existing value for labels.

Request Examples¶

{
  "name": "my-experiment-new-name",
  "environment": "python3.8"
}

{
  "name": "my-experiment-new-name",
  "description": "My new description",
  "environment": "python3.7",
  "instance_type": "4096mb",
  "default_bucket": "default",
  "labels": {
    "type": "pytorch"
  }
}

Response Structure¶

The details of the experiment

id: Unique identifier for the experiment (UUID)
name: Experiment name
description: Description of the experiment
environment: Environment of the experiment
environment_display_name: Human readable name of the environment
status: The status of the experiment
active_revision: UUID of the active revision of the experiment
latest_revision: UUID of the latest revision of the experiment
instance_type: The reserved instance type for the experiment
maximum_instances: Upper bound of number of experiment pods running in parallel
minimum_instances: Lower bound of number of experiment pods running in parallel
maximum_idle_time: Maximum time in seconds a experiment stays idle before it is stopped
labels: Dictionary containing key/value pairs where key indicates the label and value is the corresponding value of that label
creation_date: The date when the experiment was created
last_updated: The date when the experiment was last updated
monitoring: Name of a notification group which contains contacts to send notifications when runs for the experiment fail and recover
default_notification_group: Name of a notification group which contains contacts to send notifications when runs for the experiment are completed
request_retention_time: Number of seconds to store runs of the experiment
request_retention_mode: Mode of retention for runs to the experiment. It can be one of the following:
- none - the runs will not be stored
- metadata - only the metadata of the runs will be stored
- full - both the metadata and input/output of the runs will be stored
maximum_queue_size_express: Maximum number of queued express runs for all instances of this experiment
maximum_queue_size_batch: Maximum number of queued batch runs for all instances of this experiment
has_request_method: Whether the latest revision of the experiment has a 'request' method
has_requests_method: Whether the latest revision of the experiment has a 'requests' method
static_ip: A boolean indicating whether the experiment should get a static IP
restart_request_interruption: A boolean indicating whether the runs should be restarted in case of an interruption
default_bucket: Indicates in which bucket input and output artifacts of the training job will be stored

Response Examples¶

{
  "active_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "creation_date": "2021-06-15T10:12:11.554682Z",
  "default_bucket": "default",
  "default_notification_group": None,
  "description": "An experiment with Python 3.7",
  "environment": "python3-7",
  "environment_display_name": "Python 3.7",
  "has_request_method": True,
  "has_requests_method": False,
  "id": "49d7d195-2296-43c0-8eab-84b0acd6e5bc",
  "instance_type": "2048mb",
  "labels": {},
  "last_updated": 2021-06-15T10:12:11.554682Z,
  "latest_revision": "7169cac3-74eb-4189-99d7-322bc71f070b",
  "maximum_idle_time": 300,
  "maximum_instances": 10,
  "maximum_queue_size_batch": 100000,
  "maximum_queue_size_express": 100,
  "minimum_instances": 0,
  "monitoring": None,
  "name": "training-experiment",
  "request_retention_mode": "full",
  "request_retention_time": 31536000,
  "restart_request_interruption": False,
  "static_ip": False,
  "status": "available"
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
new_experiment_name = 'experiment_new_name_example' # str

# Update experiment
api_response = training.experiments_update(
    project_name=project_name,
    experiment_name=experiment_name,
    data=ubiops.ExperimentUpdate(name=new_experiment_name)
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
new_experiment_name = 'experiment_new_name_example' # str

# Update experiment
api_response = training.experiments_update(
    project_name=project_name,
    experiment_name=experiment_name,
    data=ubiops.ExperimentUpdate(name=new_experiment_name)
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
data	ExperimentUpdate

Return type¶

ExperimentDetail

Authorization¶

API token

[Back to top]

experiment_runs_create¶

ExperimentRunCreateResponse experiment_runs_create(project_name, experiment_name, data, timeout=None)

Create an experiment run

Description¶

Helper function to create an experiment run

Required Parameters¶

name: The name of the experiment run
training_code: A reference to a location containing the training code. Can be an UbiOps file uri, directory or file

Optional Parameters¶

description: Description of the experiment run
training_data: A reference to a location containing the training data. Can be an UbiOps file uri, directory or file
parameters: Dictionary containing key/value pairs with training parameters

These parameters should be given as query parameters:

timeout: Timeout for the run in seconds. The maximum allowed value is 172800 (48 hours) and the default value is 14400 (4 hours).

Request Examples¶

{
  "name": "my-run",
  "training_code": "train.py"
}

{
  "name": "my-run",
  "training_code": "ubiops-file://default/train.py",
  "training_data": "my-training-data-directory",
  "parameters = {
    "nr_epochs": 15,
    "batch_size": 32
  }
}

Response Structure¶

Details of the created run

id: Unique identifier for the run (UUID)
experiment: Experiment name for which the run was created
status: The status of the run
time_created: The date when the run was created

Response Examples¶

{
  "id": "4ae7d14b-4803-4e16-b96d-3b18caa4b605",
  "experiment": "my-experiment-name",
  "status": "pending",
  "time_created": "2020-05-12T16:23:15.456812Z"
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
timeout = 56 # int (optional)

# Create run
api_response = training.experiment_runs_create(
    project_name=project_name,
    experiment_name=experiment_name,
    data=ubiops.ExperimentRunCreate(
        name="my-run",
        training_code="./train.py"
    ),
    timeout=timeout
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
timeout = 56 # int (optional)

# Create run
api_response = training.experiment_runs_create(
    project_name=project_name,
    experiment_name=experiment_name,
    data=ubiops.ExperimentRunCreate(
        name="my-run",
        training_code="./train.py"
    ),
    timeout=timeout
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
data	ExperimentRunCreate

Return type¶

ExperimentRunCreateResponse

Authorization¶

API token

[Back to top]

experiment_runs_delete¶

experiment_runs_delete(project_name, experiment_name, run_id)

Delete an experiment run

Description¶

Helper function to delete an experiment run

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
run_id = "9eaee2b8-daf9-49fc-b958-be31f6ac48e3" # str

# Delete run
training.experiment_runs_delete(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id=run_id
)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
run_id = "9eaee2b8-daf9-49fc-b958-be31f6ac48e3" # str

# Delete run
training.experiment_runs_delete(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id=run_id,
)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
run_id	str

Return type¶

void (empty response body)

Authorization¶

API token

[Back to top]

experiment_runs_get¶

ExperimentRunDetail experiment_runs_get(project_name, experiment_name, run_id)

Get an experiment run

Description¶

Helper function to get an experiment run details

Response Structure¶

Get the details of an experiment run

id: Unique identifier for the run
experiment: Experiment name for which the run was created
status: Status of the run. Can be 'pending', 'processing', 'failed', 'completed', 'cancelled' or 'cancelled_pending'.
success: A boolean value that indicates whether the run was successful. NULL if the run is not yet finished.
time_created: Server time that the run was made (current time)
time_started: Server time that the processing of the run was started
time_completed: Server time that the processing of the run was completed
error_message: An error message explaining why the run has failed. NULL if the run was successful.
retries: Number of times that the run has been retried
request_data: A dictionary containing the data that was sent when the run was created
result: Run result. NULL if the run is 'pending', 'processing' or 'failed'.
notification_group: Name of a notification group to send notifications (e.g., emails) when the run is completed
origin: A dictionary containing the information on where the run originated from. It contains:
created_by field indicating the user that created the run

Response Examples¶

{
  "id": "2f909aeb-5c7e-4974-970d-cd0a6a073aca",
  "experiment": "my-experiment-name",
  "status": "completed",
  "success": True,
  "time_created": "2020-03-29T08:09:10.729+00:00",
  "time_started": "2020-03-28T20:00:41.276+00:00",
  "time_completed": "2020-03-28T20:00:42.241+00:00",
  "error_message": None,
  "retries": 0,
  "request_data": {
    "name": "my-experiment-run-name",
    "training_code": "ubiops-file://default/training_code.zip",
    "parameters": {
      "batch_size": 32,
      "nr_epochs": 15
    }
  },
  "result": {
    "artifact": "ubiops-file://default/a83f4004-412b-4cd9-be05-55a2789438e8/model.pkl",
    "metadata": None,
    "metrics": {"accuracy": 1.0, "loss": 0.0}
  },
  "notification_group": None,
  "origin": {
    "created_by": "[email protected]",
  }
}

{
  "id": "2f909aeb-5c7e-4974-970d-cd0a6a073aca",
  "experiment": "my-experiment-name",
  "status": "completed",
  "success": False,
  "time_created": "2020-03-29T08:09:10.729+00:00",
  "time_started": "2020-03-28T20:00:41.276+00:00",
  "time_completed": "2020-03-28T20:00:42.241+00:00",
  "error_message": "An exception occurred in the request. See the logs for details.",
  "retries": 0,
  "request_data": {
    "name": "my-experiment-run-name",
    "training_code": "ubiops-file://default/training_code.zip",
  },
  "result": None,
  "notification_group": None,
  "origin": {
    "created_by": "[email protected]",
  }
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
run_id = "9eaee2b8-daf9-49fc-b958-be31f6ac48e3" # str

# Get experiment run
api_response = training.experiment_runs_get(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id=run_id
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
run_id = "9eaee2b8-daf9-49fc-b958-be31f6ac48e3" # str

# Get experiment run
api_response = training.experiment_runs_get(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id=run_id
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
run_id	str

Return type¶

ExperimentRunDetail

Authorization¶

API token

[Back to top]

experiment_runs_list¶

list[ExperimentRunList] experiment_runs_list(project_name, experiment_name)

List experiment runs

Description¶

Helper function to list experiment runs

Response Structure¶

A list of details of the experiment runs

id: Unique identifier for the run (UUID)
experiment: Experiment name for which the run was created
status: Status of the run. Can be 'pending', 'processing', 'failed', 'completed', 'cancelled' or 'cancelled_pending'.
success: A boolean value that indicates whether the run was successful. NULL if the run is not yet finished.
time_created: Server time that the run was made (current time)
time_started: Server time that the processing of the run was started
time_completed: Server time that the processing of the run was completed

Response Examples¶

[
  {
    "id": "69eca481-8576-49e8-8e20-ea56f2005bcb",
    "experiment": "my-experiment-name",
    "status": "pending",
    "success": false,
    "time_created": "2020-03-28T20:00:26.613+00:00",
    "time_started": "2020-03-28T20:00:41.276+00:00",
    "time_completed": "2020-03-28T20:00:42.241+00:00"
  },
  {
    "id": "2521378e-263e-4e2e-85e9-a96254b36536",
    "experiment": "my-experiment-name",
    "status": "completed",
    "success": true,
    "time_created": "2020-03-28T20:00:26.613+00:00",
    "time_started": "2020-03-28T20:00:41.276+00:00",
    "time_completed": "2020-03-28T20:00:42.241+00:00"
  }
]

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# List runs
api_response = training.experiment_runs_list(
    project_name=project_name,
    experiment_name=experiment_name
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Create run
api_response = training.experiment_runs_list(
    project_name=project_name,
    experiment_name=experiment_name
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str

Return type¶

list[ExperimentRunList]

Authorization¶

API token

[Back to top]

experiment_runs_update¶

ExperimentRunUpdateResponse experiment_runs_update(project_name, experiment_name, run_id, data)

Cancel an experiment run

Description¶

Helper function to cancel an experiment run

Required Parameters¶

status: Status that the run will be updated to. It can only be cancelled.

Request Examples¶

{
  "status": "cancelled"
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# Update a run
api_response = training.experiment_runs_update(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id="9eaee2b8-daf9-49fc-b958-be31f6ac48e3",
    data=ubiops.ExperimentRunUpdate(
        status="cancelled"
    )
)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
run_id = "9eaee2b8-daf9-49fc-b958-be31f6ac48e3" # str

# Update a run
api_response = training.experiment_runs_update(
    project_name=project_name,
    experiment_name=experiment_name,
    run_id=run_id,
    data=ubiops.ExperimentRunUpdate(
        status="cancelled"
    )
)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
run_id	str
data	ExperimentRunUpdate

Return type¶

ExperimentRunUpdateResponse

Authorization¶

API token

[Back to top]

experiment_environment_variables_create¶

EnvironmentVariableList experiment_environment_variables_create(project_name, experiment_name, data)

Create experiment environment variable

Description¶

Create an environment variable for the experiment. Variables inherited from the project can be shadowed by creating a variable with the same name.

Required Parameters¶

name: The name of the variable. The variable will have this name when accessed from your training code. The variable name should contain only letters and underscores, and not start or end with an underscore.
value: The value of the variable as a string. It may be an empty string ("").
secret: If this variable contains sensitive information, set this to true to hide it from other users.

Request Examples¶

{
  "name": "experiment_variable",
  "value": "another one",
  "secret": false
}

Response Structure¶

A list of variables described by the following fields:

id: Unique identifier for the environment variable
name: Variable name
value: Variable value (will be null for secret variables)
secret: Boolean that indicates if this variable contains sensitive information

Response Examples¶

{
  "id": "54e94fbe-507e-4fae-981d-227c28a2dab0",
  "name": "experiment_variable",
  "value": "another one",
  "secret": false
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
data = ubiops.EnvironmentVariableCreate() # EnvironmentVariableCreate

# Create experiment environment variable
api_response = training.experiment_environment_variables_create(project_name, experiment_name, data)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
data = ubiops.EnvironmentVariableCreate() # EnvironmentVariableCreate

# Create experiment environment variable
api_response = training.experiment_environment_variables_create(project_name, experiment_name, data)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
data	EnvironmentVariableCreate

Return type¶

EnvironmentVariableList

Authorization¶

API token

[Back to top]

experiment_environment_variables_delete¶

experiment_environment_variables_delete(project_name, experiment_name, id)

Delete experiment environment variable

Description¶

Delete an environment variable of a experiment

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str

# Delete experiment environment variable
training.experiment_environment_variables_delete(project_name, experiment_name, id)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str

# Delete experiment environment variable
training.experiment_environment_variables_delete(project_name, experiment_name, id)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
id	str

Return type¶

void (empty response body)

Authorization¶

API token

[Back to top]

experiment_environment_variables_get¶

EnvironmentVariableList experiment_environment_variables_get(project_name, experiment_name, id)

Get experiment environment variable

Description¶

Retrieve details of a experiment environment variable. This cannot be used to retrieve details of inherited variables.

Response Structure¶

The details of the environment variable:

id: Unique identifier for the environment variable
name: Variable name
value: Variable value (will be null for secret variables)
secret: Boolean that indicates if this variable contains sensitive information

Response Examples¶

{
  "id": "4c15a27e-25ea-4be0-86c7-f4790389d061",
  "name": "database_schema",
  "value": "public",
  "secret": false
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str

# Get experiment environment variable
api_response = training.experiment_environment_variables_get(project_name, experiment_name, id)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str

# Get experiment environment variable
api_response = training.experiment_environment_variables_get(project_name, experiment_name, id)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
id	str

Return type¶

EnvironmentVariableList

Authorization¶

API token

[Back to top]

experiment_environment_variables_list¶

list[InheritedEnvironmentVariableList] experiment_environment_variables_list(project_name, experiment_name)

List experiment environment variables

Description¶

List the environment variables defined for the experiment. Includes environment variables defined at project level.

Response Structure¶

A list of variables described by the following fields:

id: Unique identifier for the environment variable
name: Variable name
value: Variable value (will be null for secret variables)
secret: Boolean that indicates if this variable contains sensitive information
inheritance_type: Type of parent object that this variable is inherited from - can be project, or null if the variable was defined for the experiment directly
inheritance_name: Name of the parent object that this variable is inherited from - will be null if the variable was defined for the experiment directly

Response Examples¶

[
  {
    "id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
    "name": "experiment_variable",
    "value": "some_value",
    "secret": false,
    "inheritance_type": null,
    "inheritance_name": null
  },
  {
    "id": "06c2c8be-507e-4fae-981d-54e94f22dab0",
    "name": "database_password",
    "value": null,
    "secret": true,
    "inheritance_type": "project",
    "inheritance_name": "project_name"
  }
]

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# List experiment environment variables
api_response = training.experiment_environment_variables_list(project_name, experiment_name)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str

# List experiment environment variables
api_response = training.experiment_environment_variables_list(project_name, experiment_name)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
dexperiment_name	str

Return type¶

list[InheritedEnvironmentVariableList]

Authorization¶

API token

[Back to top]

experiment_environment_variables_update¶

EnvironmentVariableList experiment_environment_variables_update(project_name, experiment_name, id, data)

Update experiment environment variable

Description¶

Update an environment variable for the experiment. This cannot be used to update inherited variables; to change an inherited variable for a specific experiment you can create a variable with the same name for the experiment.

Required Parameters¶

name: The name of the variable. The variable will have this name when accessed from your training code. The variable name should contain only letters and underscores, and not start or end with an underscore.
value: The value of the variable as a string. It may be an empty string ("").
secret: If this variable contains sensitive information, set this to true to hide it from other users. Can be updated from false to true, but not from true to false (i.e. secrets will stay secrets).

Request Examples¶

{
  "name": "experiment_variable",
  "value": "yet another one",
  "secret": false
}

Response Structure¶

The details of the environment variable:

id: Unique identifier for the environment variable
name: Variable name
value: Variable value (will be null for secret variables)
secret: Boolean that indicates if this variable contains sensitive information

Response Examples¶

{
  "id": "54e94fbe-507e-4fae-981d-227c28a2dab0",
  "name": "experiment_variable",
  "value": "yet another one",
  "secret": false
}

Example¶

Use system environment variables

import ubiops

# Set environment variables
# - UBIOPS_API_TOKEN: "Token <YOUR_API_TOKEN>"
# - UBIOPS_API_HOST: optional - default to "https://api.ubiops.com/v2.1"
training = ubiops.Training()

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str
data = ubiops.EnvironmentVariableCreate() # EnvironmentVariableCreate

# Update experiment environment variable
api_response = training.experiment_environment_variables_update(project_name, experiment_name, id, data)
print(api_response)

# Close the connection
training.api_client.close()

Use authorization parameters

import ubiops

configuration = ubiops.Configuration()
# Configure API token authorization
configuration.api_key['Authorization'] = "Token <YOUR_API_TOKEN>"
# Defining host is optional and default to "https://api.ubiops.com/v2.1"
configuration.host = "https://api.ubiops.com/v2.1"

api_client = ubiops.ApiClient(configuration)
training = ubiops.Training(api_client)

project_name = 'project_name_example' # str
experiment_name = 'experiment_name_example' # str
id = 'id_example' # str
data = ubiops.EnvironmentVariableCreate() # EnvironmentVariableCreate

# Update experiment environment variable
api_response = training.experiment_environment_variables_update(project_name, experiment_name, id, data)
print(api_response)

# Close the connection
api_client.close()

Parameters¶

Name	Type	Notes
project_name	str
experiment_name	str
id	str
data	EnvironmentVariableCreate

Return type¶

EnvironmentVariableList

Authorization¶

API token

[Back to top]