Documentation ¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type FunctionArgs ¶
FunctionArgs holds the arguments needed to build the url of a function, such as the function id and parallelism level
type FunctionResults ¶
type FunctionResults struct {
// contains filtered or unexported fields
}
FunctionResults holds the function id and the execution results of a function, be it a training or validation function
type FunctionTask ¶
type FunctionTask string
const ( Train FunctionTask = "train" Validation FunctionTask = "val" Init FunctionTask = "init" Inference FunctionTask = "infer" )
type TrainJob ¶
type TrainJob struct { K int // contains filtered or unexported fields }
TrainJob is each of the workers launched by the parameter server. The worker is responsible from managing the reference model, saving the intermediate accuracy/validation results in the history, and requesting/receiving new scheduling responses from the scheduler
func NewBasicJob ¶
NewBasicJob creates a job with no task provided yet. It will start the job api and wait for its task to be defined there.
This is the constructor used when deploying the jobs in separate pods
func NewTrainJob ¶
func NewTrainJob( logger *zap.Logger, task *api.TrainTask, schedulerCh chan *api.JobState, client *schedulerClient.Client) *TrainJob
NewTrainJob Creates a new TrainJob that will take care of a specific train request