API Reference
Packages
tensorstack.dev/v1beta1
Package v1beta1 contains API Schema definitions for the v1beta1 API group
Resource Types
AIStoreConfig
AIStoreConfig represents the configuration for using AIStore as a data storage for experiments.
Appears in:
| Field | Description |
|---|---|
secret string | Secret specifies the secret containing the API Key authorized to upload data to AIStore. |
folder string | Folder indicates the destination folder where the experiment data will be stored. |
AdvisorConfig
AdvisorConfig represents the configuration for the advisor algorithm and its parameters.
Appears in:
| Field | Description |
|---|---|
builtinAdvisorName string | BuiltInAdvisorName specifies the name of the built-in advisor algorithm to be used. Available options include: Hyperband, BOHB. |
classArgs string | ClassArgs defines the parameters specific to the chosen advisor algorithm. Different algorithms may require distinct parameters. |
AssessorConfig
AssessorConfig represents the configuration for the assessor algorithm and its parameters.
Appears in:
| Field | Description |
|---|---|
builtinAssessorName string | BuiltInAssessorName specifies the name of the built-in assessor algorithm to be used. Available options include: Medianstop, Curvefitting. |
classArgs string | ClassArgs defines the parameters specific to the chosen assessor algorithm. Different algorithms may require distinct parameters. |
AutoTuneExperiment
AutoTuneExperiment is the Schema for the autotune API.
Appears in:
| Field | Description |
|---|---|
apiVersion string | tensorstack.dev/apis |
kind string | AutoTuneExperiment |
metadata ObjectMeta | Refer to Kubernetes API documentation for fields of metadata. |
spec AutoTuneExperimentSpec | |
status AutoTuneExperimentStatus |
AutoTuneExperimentList
AutoTuneExperimentList contains a list of AutoTuneExperiment.
| Field | Description |
|---|---|
apiVersion string | tensorstack.dev/apis |
kind string | AutoTuneExperimentList |
metadata ListMeta | Refer to Kubernetes API documentation for fields of metadata. |
items AutoTuneExperiment array |
AutoTuneExperimentSpec
AutoTuneExperimentSpec defines the desired state of the AutoTuneExperiment.
Appears in:
| Field | Description |
|---|---|
aistore AIStoreConfig | AIStore configures an AIStore as storage for experiment data. If this field is set, data will be uploaded to the AIStore. |
maxExecSeconds integer | MaxExecSeconds is the time limit (in seconds) for the AutoTuneExperiment, If this limit is exceeded, the AutoTuneExperiment reaches phase TIMEOUT. |
maxTrialNum integer | MaxTrialNum specifies the maximum number of trials for the AutoTuneExperiment. Once this number of trials is reached, the AutoTuneExperiment reaches phase DONE. |
trialConcurrency integer | TrialConcurrency is the maximum number of trials running simultaneously. |
searchSpace string | SearchSpace defines the hyperparameter search space as a JSON string. It specifies the range for searching optimized hyperparameters. Example: { “batch_size”: {“_type”: “choice”, “_value”: [16, 32, 64, 128]}, “learning_rate”: {“_type”: “choice”, “_value”: [0.0001, 0.001, 0.01, 0.1]}, “conv_channels1”: {“_type”: “choice”, “_value”: [16, 32, 64, 128]} } |
storage Quantity | Storage specifies the size of PVC for storing configurations and training metrics. |
trainingConfig TrainingConfig | TrainingConfig represents the configuration for creating Jobs, which evaluate the performance of different hyperparameters. |
tuner TunerConfig | Tuner configures a tuner for optimizing hyperparameter. |
assessor AssessorConfig | Assessor configures an assessor for filtering hyperparameters and interrupting training when hyperparameters are deemed unqualified. Note that this field is ignored if Tuner is not set. |
advisor AdvisorConfig | Advisor configures an advisor for optimizing hyperparameter. Note that when both Tuner and Advisor are set, Tuner takes precedence. |
AutoTuneExperimentStatus
AutoTuneExperimentStatus defines the observed state of the AutoTuneExperiment.
Appears in:
| Field | Description |
|---|---|
OwnerStatus OwnerStatus | |
nextCheckedTime Time | NextCheckedTime indicates the scheduled time for the next status check of the experiment process by the controller. |
phase ExperimentStatus | Phase provides a simple, high-level summary of where the AutoTuneExperiment is in its lifecycle. Note that it is NOT intended to serve as a comprehensive state machine. This field is optional. |
serverNote string | ServerNote contains the current status of the experiment process. |
ExperimentStatus
Underlying type: string
Appears in:
TunerConfig
TunerConfig represents the configuration for the tuner algorithm and its parameters.
Appears in:
| Field | Description |
|---|---|
builtinTunerName string | BuiltInTunerName specifies the name of the built-in tuner algorithm to be used. Available options include: Random, Anneal, TPE, Evolution, Batch, GridSearch, MetisTuner, GPTuner, PPOTuner, PBTTuner. |
classArgs string | ClassArgs defines the parameters specific to the chosen tuner algorithm. Different algorithms may require distinct parameters. |