Models¶
Warning
The Models client is in an Alpha release. Descartes Labs reserves the right to modify this API without any guarantees about backwards compatibility.
Models |
Models is a client for interacting with registered models and tracking the status of jobs. |
-
class
Models
(host=None, port=None, auth=None, certificate=None, default_retry=None, default_metadata=None, use_insecure_channel=False)[source]¶ Models is a client for interacting with registered models and tracking the status of jobs.
Parameters: - host (str, optional) – The backend host to connect to. Defaults to the correct value. Only override when debugging.
- port (int, optional) – The backend port to connect to. Defaults to the correct value. Only override when debugging.
- auth (Auth, optional) – The authentication instance to use. Defaults to
Auth.get_default_client()
. - certificate (bytes, optional) – The certificate to use when connecting to the backend service.
- default_retry (Retry, int, optional) – The retry instance or number of times to retry a connection before giving up. Default is to retry 5 times for retryable status codes.
- default_metadata (tuple, optional) – Metadata (headers) to send to every RPC when called. The default is empty.
- use_insecure_channel (bool, optional) – If set, an insecure channel will be used. Default is secure; only use insecure when debugging.
Attributes:
DEFAULT_TIMEOUT
STREAM_TIMEOUT
api
The available Client operations, as a dict. certificate
The Client SSL certificate. channel
The GRPC channel of the Client. token
The Client token. Methods:
INSECURE_CHANNEL_FACTORY
(target[, options, …])Creates an insecure Channel to a server. SECURE_CHANNEL_FACTORY
(target, credentials)Creates a secure Channel to a server. close
()Close the GRPC channel associated with the Client. get_default_client
()Retrieve the default client. get_model
(model_name)Gets ModelSpec from model name. get_status
(work_id)Gets the status of a job health
([timeout])Check the health of the GRPC server (SERVING, NOT_SERVING, UNKNOWN). invoke
(model_name, parameters)Invokes a model model_name
with specified tuning parameters.set_default_client
(client)Change the default client to the given client. -
static
INSECURE_CHANNEL_FACTORY
(target, options=None, compression=None)¶ Creates an insecure Channel to a server.
The returned Channel is thread-safe.
Parameters: - target – The server address
- options – An optional list of key-value pairs (channel_arguments in gRPC Core runtime) to configure the channel.
- compression – An optional value indicating the compression method to be used over the lifetime of the channel. This is an EXPERIMENTAL option.
Returns: A Channel.
-
static
SECURE_CHANNEL_FACTORY
(target, credentials, options=None, compression=None)¶ Creates a secure Channel to a server.
The returned Channel is thread-safe.
Parameters: - target – The server address.
- credentials – A ChannelCredentials instance.
- options – An optional list of key-value pairs (channel_arguments in gRPC Core runtime) to configure the channel.
- compression – An optional value indicating the compression method to be used over the lifetime of the channel. This is an EXPERIMENTAL option.
Returns: A Channel.
-
close
()¶ Close the GRPC channel associated with the Client.
-
classmethod
get_default_client
()¶ Retrieve the default client.
This client is used whenever you don’t explicitly set the client.
-
get_model
(model_name: str) → descarteslabs.common.services.currier.spec.internal.ModelSpec[source]¶ Gets ModelSpec from model name.
Parameters: model_name (str) – Name of model to fetch. Returns: model_spec – Deserialized ModelSpec object. Return type: ModelSpec
-
get_status
(work_id: str) → Dict[str, float][source]¶ Gets the status of a job
Parameters: work_id (str) – Work ID of the job to check the status on. Returns: status – Percentages for work pending, running, cancelled, successful and failed. Return type: Dict[str, float]
-
health
(timeout=None)¶ Check the health of the GRPC server (SERVING, NOT_SERVING, UNKNOWN).
Example
>>> from descarteslabs.client.grpc import GrpcClient >>> GrpcClient().health() SERVING
-
invoke
(model_name: str, parameters: Dict[str, Any]) → str[source]¶ Invokes a model
model_name
with specified tuning parameters.Parameters: Returns: work_id – Work ID of the invoked job. This can be used for checking the status of the work with
get_status
.Return type: str
-
classmethod
set_default_client
(client)¶ Change the default client to the given client.
This is the client that will be used whenever you don’t explicitly set the client
-
DEFAULT_TIMEOUT
= 5¶
-
STREAM_TIMEOUT
= 86400¶
-
property
api
¶ The available Client operations, as a dict.
-
property
certificate
¶ The Client SSL certificate.
-
property
channel
¶ The GRPC channel of the Client.
-
property
token
¶ The Client token.