yr.InvokeOptions

class yr.InvokeOptions(cpu: int = 500, memory: int = 500, concurrency: int | None = None, custom_resources: ~typing.Dict[str, float] = <factory>, custom_extensions: ~typing.Dict[str, str] = <factory>, pod_labels: ~typing.Dict[str, str] = <factory>, labels: ~typing.List[str] = <factory>, affinity: ~typing.Dict[str, str] = <factory>, device: ~yr.config.Device = <factory>, max_invoke_latency: int = 5000, min_instances: int = 0, max_instances: int = 0, recover_retry_times: int = 0, need_order: bool = False, name: str = '', namespace: str = '', schedule_affinities: ~typing.List[~yr.affinity.Affinity] = <factory>, is_data_affinity: bool = False, resource_group_options: ~yr.config.ResourceGroupOptions = <factory>, function_group_options: ~yr.config.FunctionGroupOptions = <factory>, env_vars: ~typing.Dict[str, str] = <factory>, retry_times: int = 0, trace_id: str = '', alias_params: ~typing.Dict[str, str] = <factory>, runtime_env: ~typing.Dict = <factory>, debug: ~yr.config.DebugConfig = <factory>, preempted_allowed: bool = False, instance_priority: int = 0, schedule_timeout_ms: int = 30000, group_name: str = '', is_delete_remote_tensor: bool = False, get_if_exists: bool = False, skip_serialize: bool = False, port_forwardings: ~typing.List[~yr.config.PortForwarding] = <factory>)

Bases: object

Use to set the invoke options.

Examples

>>> import yr
>>> import time
>>> yr.init()
>>> opt = yr.InvokeOptions()
>>> opt.pod_labels["k1"] = "v1"
>>> @yr.invoke(invoke_options=opt)
... def func():
...     time.sleep(100)
>>> ret = func.invoke()
>>> yr.get(ret)
>>> yr.finalize()

Attributes

concurrency	Instance concurrency.
cpu	The size of the CPU required.
get_if_exists
group_name
instance_priority	The priority of an instance is determined by its value.
is_data_affinity	Whether to enable data affinity scheduling.
is_delete_remote_tensor
max_instances	Specify the maximum number of instances for a stateless function.
max_invoke_latency	Specify the time when the invoke call of the desired heterogeneous function is completed.
memory	[128, 1073741824].
min_instances	Specify the minimum number of instances for a stateless function.
name	Used to specify the name of the instance.
namespace	Used to specify the namespace of the instance.
need_order	Whether to enable order-preserving.
preempted_allowed	Whether an instance can be preempted is effective only in the priority scenario (when the maxPriority configuration item deployed by YuanRong is greater than 0).
preferred_anti_other_labels	Whether to enable anti-affinity for non-selectable resources.
preferred_priority	Set whether to enable weak affinity priority scheduling.
recover_retry_times	The number of instance recovery times (when an instance abnormally exits, the instance is automatically restored to the latest state).
required_priority	Set whether to enable strong affinity priority scheduling.
retry_times	Number of retries for stateless functions.
schedule_timeout_ms	milliseconds.
skip_serialize	Whether to skip serializing the instance class code.
trace_id	Set the traceId for function calls for link tracing.
custom_resources	Custom resources currently support "GPU/XX/YY" and "NPU/XX/YY", where XX is the card model such as Ascend910B4, and YY can be count, latency, or stream.
custom_extensions	Specify user-defined configurations, such as function concurrency.
pod_labels	Pod labels only used in Kubernetes environment.
labels	Labels of instance
affinity	Affinity of instance
device	Specify the name of the model used by the heterogeneous function.
schedule_affinities	Set affinity condition list.
resource_group_options	Specify the ResourceGroup option, which includes resource_group_name and bundle_index.
function_group_options	Function group options.
env_vars	Set environment variables when the instance starts.
alias_params	In the scenario where a function is invoked by a specified alias in cross-function invocation, when the alias is a rule alias, this parameter is used to set the kv parameter that the rule alias depends on.
runtime_env
debug	Configure the stateful/stateless function runtime environment with conda, pip, working_dir, and env_vars.
port_forwardings	Configure port forwarding rules for the sandbox.

Methods

__init__
check_options_range	Check whether the options are in the valid range.
check_options_valid	Check whether the options are valid.