Documentation ¶
Index ¶
- Variables
- type CheckpointWorkload
- func (*CheckpointWorkload) Descriptor() ([]byte, []int)deprecated
- func (x *CheckpointWorkload) GetEndTime() *timestamp.Timestamp
- func (x *CheckpointWorkload) GetMetadata() *_struct.Struct
- func (x *CheckpointWorkload) GetResources() map[string]int64
- func (x *CheckpointWorkload) GetState() checkpointv1.State
- func (x *CheckpointWorkload) GetTotalBatches() int32
- func (x *CheckpointWorkload) GetUuid() string
- func (*CheckpointWorkload) ProtoMessage()
- func (x *CheckpointWorkload) ProtoReflect() protoreflect.Message
- func (x *CheckpointWorkload) Reset()
- func (x *CheckpointWorkload) String() string
- type MetricsReport
- func (*MetricsReport) Descriptor() ([]byte, []int)deprecated
- func (x *MetricsReport) GetArchived() bool
- func (x *MetricsReport) GetEndTime() *timestamp.Timestamp
- func (x *MetricsReport) GetGroup() string
- func (x *MetricsReport) GetId() int32
- func (x *MetricsReport) GetMetrics() *_struct.Struct
- func (x *MetricsReport) GetTotalBatches() int32
- func (x *MetricsReport) GetTrialId() int32
- func (x *MetricsReport) GetTrialRunId() int32
- func (*MetricsReport) ProtoMessage()
- func (x *MetricsReport) ProtoReflect() protoreflect.Message
- func (x *MetricsReport) Reset()
- func (x *MetricsReport) String() string
- type MetricsWorkload
- func (*MetricsWorkload) Descriptor() ([]byte, []int)deprecated
- func (x *MetricsWorkload) GetEndTime() *timestamp.Timestamp
- func (x *MetricsWorkload) GetMetrics() *commonv1.Metrics
- func (x *MetricsWorkload) GetNumInputs() int32deprecated
- func (x *MetricsWorkload) GetTotalBatches() int32
- func (*MetricsWorkload) ProtoMessage()
- func (x *MetricsWorkload) ProtoReflect() protoreflect.Message
- func (x *MetricsWorkload) Reset()
- func (x *MetricsWorkload) String() string
- type RendezvousInfo
- func (*RendezvousInfo) Descriptor() ([]byte, []int)deprecated
- func (x *RendezvousInfo) GetAddresses() []string
- func (x *RendezvousInfo) GetRank() int32
- func (x *RendezvousInfo) GetSlots() []int32
- func (*RendezvousInfo) ProtoMessage()
- func (x *RendezvousInfo) ProtoReflect() protoreflect.Message
- func (x *RendezvousInfo) Reset()
- func (x *RendezvousInfo) String() string
- type State
- type Trial
- func (*Trial) Descriptor() ([]byte, []int)deprecated
- func (x *Trial) GetBestCheckpoint() *CheckpointWorkload
- func (x *Trial) GetBestValidation() *MetricsWorkload
- func (x *Trial) GetCheckpointCount() int32
- func (x *Trial) GetEndTime() *timestamp.Timestamp
- func (x *Trial) GetExperimentId() int32
- func (x *Trial) GetHparams() *_struct.Struct
- func (x *Trial) GetId() int32
- func (x *Trial) GetLatestValidation() *MetricsWorkload
- func (x *Trial) GetLogRetentionDays() int32
- func (x *Trial) GetMetadata() *_struct.Struct
- func (x *Trial) GetRestarts() int32
- func (x *Trial) GetRunnerState() string
- func (x *Trial) GetSearcherMetricValue() float64
- func (x *Trial) GetStartTime() *timestamp.Timestamp
- func (x *Trial) GetState() State
- func (x *Trial) GetSummaryMetrics() *_struct.Struct
- func (x *Trial) GetTaskId() stringdeprecated
- func (x *Trial) GetTaskIds() []string
- func (x *Trial) GetTotalBatchesProcessed() int32
- func (x *Trial) GetTotalCheckpointSize() uint64
- func (x *Trial) GetWallClockTime() float64
- func (x *Trial) GetWarmStartCheckpointUuid() string
- func (*Trial) ProtoMessage()
- func (x *Trial) ProtoReflect() protoreflect.Message
- func (x *Trial) Reset()
- func (x *Trial) String() string
- type TrialEarlyExit
- type TrialEarlyExit_ExitedReason
- func (TrialEarlyExit_ExitedReason) Descriptor() protoreflect.EnumDescriptor
- func (x TrialEarlyExit_ExitedReason) Enum() *TrialEarlyExit_ExitedReason
- func (TrialEarlyExit_ExitedReason) EnumDescriptor() ([]byte, []int)deprecated
- func (x TrialEarlyExit_ExitedReason) Number() protoreflect.EnumNumber
- func (x TrialEarlyExit_ExitedReason) String() string
- func (TrialEarlyExit_ExitedReason) Type() protoreflect.EnumType
- type TrialMetrics
- func (*TrialMetrics) Descriptor() ([]byte, []int)deprecated
- func (x *TrialMetrics) GetMetrics() *commonv1.Metrics
- func (x *TrialMetrics) GetReportTime() *timestamp.Timestamp
- func (x *TrialMetrics) GetStepsCompleted() int32
- func (x *TrialMetrics) GetTrialId() int32
- func (x *TrialMetrics) GetTrialRunId() int32
- func (*TrialMetrics) ProtoMessage()
- func (x *TrialMetrics) ProtoReflect() protoreflect.Message
- func (x *TrialMetrics) Reset()
- func (x *TrialMetrics) String() string
- type TrialProfilerMetricLabels
- func (*TrialProfilerMetricLabels) Descriptor() ([]byte, []int)deprecated
- func (x *TrialProfilerMetricLabels) GetAgentId() string
- func (x *TrialProfilerMetricLabels) GetGpuUuid() string
- func (x *TrialProfilerMetricLabels) GetMetricType() TrialProfilerMetricLabels_ProfilerMetricType
- func (x *TrialProfilerMetricLabels) GetName() string
- func (x *TrialProfilerMetricLabels) GetTrialId() int32
- func (*TrialProfilerMetricLabels) ProtoMessage()
- func (x *TrialProfilerMetricLabels) ProtoReflect() protoreflect.Message
- func (x *TrialProfilerMetricLabels) Reset()
- func (x *TrialProfilerMetricLabels) String() string
- type TrialProfilerMetricLabels_ProfilerMetricType
- func (TrialProfilerMetricLabels_ProfilerMetricType) Descriptor() protoreflect.EnumDescriptor
- func (x TrialProfilerMetricLabels_ProfilerMetricType) Enum() *TrialProfilerMetricLabels_ProfilerMetricType
- func (TrialProfilerMetricLabels_ProfilerMetricType) EnumDescriptor() ([]byte, []int)deprecated
- func (x TrialProfilerMetricLabels_ProfilerMetricType) Number() protoreflect.EnumNumber
- func (x TrialProfilerMetricLabels_ProfilerMetricType) String() string
- func (TrialProfilerMetricLabels_ProfilerMetricType) Type() protoreflect.EnumType
- type TrialProfilerMetricsBatch
- func (*TrialProfilerMetricsBatch) Descriptor() ([]byte, []int)deprecated
- func (x *TrialProfilerMetricsBatch) GetBatches() []int32
- func (x *TrialProfilerMetricsBatch) GetLabels() *TrialProfilerMetricLabels
- func (x *TrialProfilerMetricsBatch) GetTimestamps() []*timestamp.Timestamp
- func (x *TrialProfilerMetricsBatch) GetValues() []float32
- func (*TrialProfilerMetricsBatch) ProtoMessage()
- func (x *TrialProfilerMetricsBatch) ProtoReflect() protoreflect.Message
- func (x *TrialProfilerMetricsBatch) Reset()
- func (x *TrialProfilerMetricsBatch) String() string
- type TrialRunnerMetadata
- func (*TrialRunnerMetadata) Descriptor() ([]byte, []int)deprecated
- func (x *TrialRunnerMetadata) GetState() string
- func (*TrialRunnerMetadata) ProtoMessage()
- func (x *TrialRunnerMetadata) ProtoReflect() protoreflect.Message
- func (x *TrialRunnerMetadata) Reset()
- func (x *TrialRunnerMetadata) String() string
- type TrialSourceInfo
- func (*TrialSourceInfo) Descriptor() ([]byte, []int)deprecated
- func (x *TrialSourceInfo) GetCheckpointUuid() string
- func (x *TrialSourceInfo) GetModelId() int32
- func (x *TrialSourceInfo) GetModelVersion() int32
- func (x *TrialSourceInfo) GetTrialId() int32
- func (x *TrialSourceInfo) GetTrialSourceInfoType() TrialSourceInfoType
- func (*TrialSourceInfo) ProtoMessage()
- func (x *TrialSourceInfo) ProtoReflect() protoreflect.Message
- func (x *TrialSourceInfo) Reset()
- func (x *TrialSourceInfo) String() string
- type TrialSourceInfoType
- func (TrialSourceInfoType) Descriptor() protoreflect.EnumDescriptor
- func (x TrialSourceInfoType) Enum() *TrialSourceInfoType
- func (TrialSourceInfoType) EnumDescriptor() ([]byte, []int)deprecated
- func (x TrialSourceInfoType) Number() protoreflect.EnumNumber
- func (x TrialSourceInfoType) String() string
- func (TrialSourceInfoType) Type() protoreflect.EnumType
Constants ¶
This section is empty.
Variables ¶
var ( State_name = map[int32]string{ 0: "STATE_UNSPECIFIED", 1: "STATE_ACTIVE", 2: "STATE_PAUSED", 3: "STATE_STOPPING_CANCELED", 4: "STATE_STOPPING_KILLED", 5: "STATE_STOPPING_COMPLETED", 6: "STATE_STOPPING_ERROR", 7: "STATE_CANCELED", 8: "STATE_COMPLETED", 9: "STATE_ERROR", 10: "STATE_QUEUED", 11: "STATE_PULLING", 12: "STATE_STARTING", 13: "STATE_RUNNING", } State_value = map[string]int32{ "STATE_UNSPECIFIED": 0, "STATE_ACTIVE": 1, "STATE_PAUSED": 2, "STATE_STOPPING_CANCELED": 3, "STATE_STOPPING_KILLED": 4, "STATE_STOPPING_COMPLETED": 5, "STATE_STOPPING_ERROR": 6, "STATE_CANCELED": 7, "STATE_COMPLETED": 8, "STATE_ERROR": 9, "STATE_QUEUED": 10, "STATE_PULLING": 11, "STATE_STARTING": 12, "STATE_RUNNING": 13, } )
Enum value maps for State.
var ( TrialSourceInfoType_name = map[int32]string{ 0: "TRIAL_SOURCE_INFO_TYPE_UNSPECIFIED", 1: "TRIAL_SOURCE_INFO_TYPE_INFERENCE", 2: "TRIAL_SOURCE_INFO_TYPE_FINE_TUNING", } TrialSourceInfoType_value = map[string]int32{ "TRIAL_SOURCE_INFO_TYPE_UNSPECIFIED": 0, "TRIAL_SOURCE_INFO_TYPE_INFERENCE": 1, "TRIAL_SOURCE_INFO_TYPE_FINE_TUNING": 2, } )
Enum value maps for TrialSourceInfoType.
var ( TrialProfilerMetricLabels_ProfilerMetricType_name = map[int32]string{ 0: "PROFILER_METRIC_TYPE_UNSPECIFIED", 1: "PROFILER_METRIC_TYPE_SYSTEM", 2: "PROFILER_METRIC_TYPE_TIMING", 3: "PROFILER_METRIC_TYPE_MISC", } TrialProfilerMetricLabels_ProfilerMetricType_value = map[string]int32{ "PROFILER_METRIC_TYPE_UNSPECIFIED": 0, "PROFILER_METRIC_TYPE_SYSTEM": 1, "PROFILER_METRIC_TYPE_TIMING": 2, "PROFILER_METRIC_TYPE_MISC": 3, } )
Enum value maps for TrialProfilerMetricLabels_ProfilerMetricType.
var ( TrialEarlyExit_ExitedReason_name = map[int32]string{ 0: "EXITED_REASON_UNSPECIFIED", 1: "EXITED_REASON_INVALID_HP", 3: "EXITED_REASON_INIT_INVALID_HP", } TrialEarlyExit_ExitedReason_value = map[string]int32{ "EXITED_REASON_UNSPECIFIED": 0, "EXITED_REASON_INVALID_HP": 1, "EXITED_REASON_INIT_INVALID_HP": 3, } )
Enum value maps for TrialEarlyExit_ExitedReason.
var File_determined_trial_v1_trial_proto protoreflect.FileDescriptor
Functions ¶
This section is empty.
Types ¶
type CheckpointWorkload ¶
type CheckpointWorkload struct { // UUID of the checkpoint. Uuid string `protobuf:"bytes,1,opt,name=uuid,proto3" json:"uuid,omitempty"` // The time the workload finished or was stopped. EndTime *timestamp.Timestamp `protobuf:"bytes,3,opt,name=end_time,json=endTime,proto3" json:"end_time,omitempty"` // The state of the checkpoint. State checkpointv1.State `protobuf:"varint,4,opt,name=state,proto3,enum=determined.checkpoint.v1.State" json:"state,omitempty"` // Dictionary of file paths to file sizes in bytes of all files in the // checkpoint. Resources map[string]int64 `` /* 160-byte string literal not displayed */ // Total number of batches as of this workload's completion. TotalBatches int32 `protobuf:"varint,8,opt,name=total_batches,json=totalBatches,proto3" json:"total_batches,omitempty"` // User defined metadata associated with the checkpoint. Metadata *_struct.Struct `protobuf:"bytes,9,opt,name=metadata,proto3" json:"metadata,omitempty"` // contains filtered or unexported fields }
CheckpointWorkload is an artifact created by a trial during training.
func (*CheckpointWorkload) Descriptor
deprecated
func (*CheckpointWorkload) Descriptor() ([]byte, []int)
Deprecated: Use CheckpointWorkload.ProtoReflect.Descriptor instead.
func (*CheckpointWorkload) GetEndTime ¶
func (x *CheckpointWorkload) GetEndTime() *timestamp.Timestamp
func (*CheckpointWorkload) GetMetadata ¶
func (x *CheckpointWorkload) GetMetadata() *_struct.Struct
func (*CheckpointWorkload) GetResources ¶
func (x *CheckpointWorkload) GetResources() map[string]int64
func (*CheckpointWorkload) GetState ¶
func (x *CheckpointWorkload) GetState() checkpointv1.State
func (*CheckpointWorkload) GetTotalBatches ¶
func (x *CheckpointWorkload) GetTotalBatches() int32
func (*CheckpointWorkload) GetUuid ¶
func (x *CheckpointWorkload) GetUuid() string
func (*CheckpointWorkload) ProtoMessage ¶
func (*CheckpointWorkload) ProtoMessage()
func (*CheckpointWorkload) ProtoReflect ¶
func (x *CheckpointWorkload) ProtoReflect() protoreflect.Message
func (*CheckpointWorkload) Reset ¶
func (x *CheckpointWorkload) Reset()
func (*CheckpointWorkload) String ¶
func (x *CheckpointWorkload) String() string
type MetricsReport ¶
type MetricsReport struct { // ID of the trial. TrialId int32 `protobuf:"varint,1,opt,name=trial_id,json=trialId,proto3" json:"trial_id,omitempty"` // End time of when metric was reported. EndTime *timestamp.Timestamp `protobuf:"bytes,2,opt,name=end_time,json=endTime,proto3" json:"end_time,omitempty"` // Struct of the reported metrics. Metrics *_struct.Struct `protobuf:"bytes,3,opt,name=metrics,proto3" json:"metrics,omitempty"` // batches completed in the report. TotalBatches int32 `protobuf:"varint,4,opt,name=total_batches,json=totalBatches,proto3" json:"total_batches,omitempty"` // If metric is archived. Archived bool `protobuf:"varint,5,opt,name=archived,proto3" json:"archived,omitempty"` // ID of metric in table. Id int32 `protobuf:"varint,6,opt,name=id,proto3" json:"id,omitempty"` // Run ID of trial when metric was reported. TrialRunId int32 `protobuf:"varint,7,opt,name=trial_run_id,json=trialRunId,proto3" json:"trial_run_id,omitempty"` // Name of the Metric Group ("training", "validation", anything else) Group string `protobuf:"bytes,8,opt,name=group,proto3" json:"group,omitempty"` // contains filtered or unexported fields }
Metrics report.
func (*MetricsReport) Descriptor
deprecated
func (*MetricsReport) Descriptor() ([]byte, []int)
Deprecated: Use MetricsReport.ProtoReflect.Descriptor instead.
func (*MetricsReport) GetArchived ¶
func (x *MetricsReport) GetArchived() bool
func (*MetricsReport) GetEndTime ¶
func (x *MetricsReport) GetEndTime() *timestamp.Timestamp
func (*MetricsReport) GetGroup ¶
func (x *MetricsReport) GetGroup() string
func (*MetricsReport) GetId ¶
func (x *MetricsReport) GetId() int32
func (*MetricsReport) GetMetrics ¶
func (x *MetricsReport) GetMetrics() *_struct.Struct
func (*MetricsReport) GetTotalBatches ¶
func (x *MetricsReport) GetTotalBatches() int32
func (*MetricsReport) GetTrialId ¶
func (x *MetricsReport) GetTrialId() int32
func (*MetricsReport) GetTrialRunId ¶
func (x *MetricsReport) GetTrialRunId() int32
func (*MetricsReport) ProtoMessage ¶
func (*MetricsReport) ProtoMessage()
func (*MetricsReport) ProtoReflect ¶
func (x *MetricsReport) ProtoReflect() protoreflect.Message
func (*MetricsReport) Reset ¶
func (x *MetricsReport) Reset()
func (*MetricsReport) String ¶
func (x *MetricsReport) String() string
type MetricsWorkload ¶
type MetricsWorkload struct { // The time the workload finished or was stopped. EndTime *timestamp.Timestamp `protobuf:"bytes,2,opt,name=end_time,json=endTime,proto3" json:"end_time,omitempty"` // Metrics. Metrics *commonv1.Metrics `protobuf:"bytes,40,opt,name=metrics,proto3" json:"metrics,omitempty"` // Number of inputs processed. // // Deprecated: Do not use. NumInputs int32 `protobuf:"varint,5,opt,name=num_inputs,json=numInputs,proto3" json:"num_inputs,omitempty"` // Total number of batches as of this workload's completion. TotalBatches int32 `protobuf:"varint,8,opt,name=total_batches,json=totalBatches,proto3" json:"total_batches,omitempty"` // contains filtered or unexported fields }
MetricsWorkload is a workload generating metrics.
func (*MetricsWorkload) Descriptor
deprecated
func (*MetricsWorkload) Descriptor() ([]byte, []int)
Deprecated: Use MetricsWorkload.ProtoReflect.Descriptor instead.
func (*MetricsWorkload) GetEndTime ¶
func (x *MetricsWorkload) GetEndTime() *timestamp.Timestamp
func (*MetricsWorkload) GetMetrics ¶
func (x *MetricsWorkload) GetMetrics() *commonv1.Metrics
func (*MetricsWorkload) GetNumInputs
deprecated
func (x *MetricsWorkload) GetNumInputs() int32
Deprecated: Do not use.
func (*MetricsWorkload) GetTotalBatches ¶
func (x *MetricsWorkload) GetTotalBatches() int32
func (*MetricsWorkload) ProtoMessage ¶
func (*MetricsWorkload) ProtoMessage()
func (*MetricsWorkload) ProtoReflect ¶
func (x *MetricsWorkload) ProtoReflect() protoreflect.Message
func (*MetricsWorkload) Reset ¶
func (x *MetricsWorkload) Reset()
func (*MetricsWorkload) String ¶
func (x *MetricsWorkload) String() string
type RendezvousInfo ¶
type RendezvousInfo struct { // The rendezvous addresses of the other containers. Addresses []string `protobuf:"bytes,1,rep,name=addresses,proto3" json:"addresses,omitempty"` // The container rank. Rank int32 `protobuf:"varint,2,opt,name=rank,proto3" json:"rank,omitempty"` // The slots for each address, respectively. Slots []int32 `protobuf:"varint,3,rep,packed,name=slots,proto3" json:"slots,omitempty"` // contains filtered or unexported fields }
The rendezvous info for the trial to rendezvous with sibling containers.
func (*RendezvousInfo) Descriptor
deprecated
func (*RendezvousInfo) Descriptor() ([]byte, []int)
Deprecated: Use RendezvousInfo.ProtoReflect.Descriptor instead.
func (*RendezvousInfo) GetAddresses ¶
func (x *RendezvousInfo) GetAddresses() []string
func (*RendezvousInfo) GetRank ¶
func (x *RendezvousInfo) GetRank() int32
func (*RendezvousInfo) GetSlots ¶
func (x *RendezvousInfo) GetSlots() []int32
func (*RendezvousInfo) ProtoMessage ¶
func (*RendezvousInfo) ProtoMessage()
func (*RendezvousInfo) ProtoReflect ¶
func (x *RendezvousInfo) ProtoReflect() protoreflect.Message
func (*RendezvousInfo) Reset ¶
func (x *RendezvousInfo) Reset()
func (*RendezvousInfo) String ¶
func (x *RendezvousInfo) String() string
type State ¶
type State int32
The current state of the trial. see \dT+ trial_state in db
const ( // The trial is in an unspecified state. State_STATE_UNSPECIFIED State = 0 // The trial is in an active state. State_STATE_ACTIVE State = 1 // The trial is in a paused state State_STATE_PAUSED State = 2 // The trial is canceled and is shutting down. State_STATE_STOPPING_CANCELED State = 3 // The trial is killed and is shutting down. State_STATE_STOPPING_KILLED State = 4 // The trial is completed and is shutting down. State_STATE_STOPPING_COMPLETED State = 5 // The trial is errored and is shutting down. State_STATE_STOPPING_ERROR State = 6 // The trial is canceled and is shut down. State_STATE_CANCELED State = 7 // The trial is completed and is shut down. State_STATE_COMPLETED State = 8 // The trial is errored and is shut down. State_STATE_ERROR State = 9 // The trial is queued (waiting to be run, or job state is still queued). // Queued is a substate of the Active state. State_STATE_QUEUED State = 10 // The trial is pulling the image. Pulling is a substate of the Active // state. State_STATE_PULLING State = 11 // The trial is preparing the environment after finishing pulling the // image. Starting is a substate of the Active state. State_STATE_STARTING State = 12 // The trial's allocation is actively running. // Running is a substate of the Active state. State_STATE_RUNNING State = 13 )
func (State) Descriptor ¶
func (State) Descriptor() protoreflect.EnumDescriptor
func (State) EnumDescriptor
deprecated
func (State) Number ¶
func (x State) Number() protoreflect.EnumNumber
func (State) Type ¶
func (State) Type() protoreflect.EnumType
type Trial ¶
type Trial struct { // The id of the trial. Id int32 `protobuf:"varint,1,opt,name=id,proto3" json:"id,omitempty"` // The id of the parent experiment. ExperimentId int32 `protobuf:"varint,2,opt,name=experiment_id,json=experimentId,proto3" json:"experiment_id,omitempty"` // The time the trial was started. StartTime *timestamp.Timestamp `protobuf:"bytes,3,opt,name=start_time,json=startTime,proto3" json:"start_time,omitempty"` // The time the trial ended if the trial is stopped. EndTime *timestamp.Timestamp `protobuf:"bytes,4,opt,name=end_time,json=endTime,proto3" json:"end_time,omitempty"` // The current state of the trial. State State `protobuf:"varint,5,opt,name=state,proto3,enum=determined.trial.v1.State" json:"state,omitempty"` // Number times the trial restarted. Restarts int32 `protobuf:"varint,17,opt,name=restarts,proto3" json:"restarts,omitempty"` // Trial hyperparameters. Hparams *_struct.Struct `protobuf:"bytes,6,opt,name=hparams,proto3" json:"hparams,omitempty"` // The current processed batches. TotalBatchesProcessed int32 `` /* 127-byte string literal not displayed */ // Best validation. BestValidation *MetricsWorkload `protobuf:"bytes,8,opt,name=best_validation,json=bestValidation,proto3" json:"best_validation,omitempty"` // Latest validation. LatestValidation *MetricsWorkload `protobuf:"bytes,9,opt,name=latest_validation,json=latestValidation,proto3" json:"latest_validation,omitempty"` // Best checkpoint. BestCheckpoint *CheckpointWorkload `protobuf:"bytes,10,opt,name=best_checkpoint,json=bestCheckpoint,proto3" json:"best_checkpoint,omitempty"` // The last reported state of the trial runner (harness code). RunnerState string `protobuf:"bytes,11,opt,name=runner_state,json=runnerState,proto3" json:"runner_state,omitempty"` // The wall clock time is all active time of the cluster for the trial, // inclusive of everything (restarts, initiailization, etc), in seconds. WallClockTime float64 `protobuf:"fixed64,12,opt,name=wall_clock_time,json=wallClockTime,proto3" json:"wall_clock_time,omitempty"` // UUID of checkpoint that this trial started from. WarmStartCheckpointUuid string `` /* 135-byte string literal not displayed */ // Id of the first task associated with this trial. // This field is deprecated since trials can have multiple tasks. // // Deprecated: Do not use. TaskId string `protobuf:"bytes,14,opt,name=task_id,json=taskId,proto3" json:"task_id,omitempty"` // The sum of sizes of all resources in all checkpoints for the trial. TotalCheckpointSize uint64 `protobuf:"varint,15,opt,name=total_checkpoint_size,json=totalCheckpointSize,proto3" json:"total_checkpoint_size,omitempty"` // The count of checkpoints. CheckpointCount int32 `protobuf:"varint,18,opt,name=checkpoint_count,json=checkpointCount,proto3" json:"checkpoint_count,omitempty"` // summary metrics SummaryMetrics *_struct.Struct `protobuf:"bytes,19,opt,name=summary_metrics,json=summaryMetrics,proto3" json:"summary_metrics,omitempty"` // Task IDs of tasks associated with this trial. // Length of task_ids will always be greater or equal to one when TaskID is // sent. For example CompareTrial we will send a reduced Trial object, without // TaskID or TaskIDs fileld in. The first element of task_ids will be the same // as task_id. task_ids is sorted ascending by task_run_id. TaskIds []string `protobuf:"bytes,20,rep,name=task_ids,json=taskIds,proto3" json:"task_ids,omitempty"` // Signed searcher metrics value. SearcherMetricValue float64 `protobuf:"fixed64,21,opt,name=searcher_metric_value,json=searcherMetricValue,proto3" json:"searcher_metric_value,omitempty"` // Number of days to retain logs for. LogRetentionDays *int32 `protobuf:"varint,22,opt,name=log_retention_days,json=logRetentionDays,proto3,oneof" json:"log_retention_days,omitempty"` // metadata associated with the trial (based off the metadata stored in the // run). Metadata *_struct.Struct `protobuf:"bytes,23,opt,name=metadata,proto3,oneof" json:"metadata,omitempty"` // contains filtered or unexported fields }
Trial is a set of workloads and are exploring a determined set of hyperparameters.
func (*Trial) Descriptor
deprecated
func (*Trial) GetBestCheckpoint ¶
func (x *Trial) GetBestCheckpoint() *CheckpointWorkload
func (*Trial) GetBestValidation ¶
func (x *Trial) GetBestValidation() *MetricsWorkload
func (*Trial) GetCheckpointCount ¶
func (*Trial) GetEndTime ¶
func (*Trial) GetExperimentId ¶
func (*Trial) GetHparams ¶
func (*Trial) GetLatestValidation ¶
func (x *Trial) GetLatestValidation() *MetricsWorkload
func (*Trial) GetLogRetentionDays ¶
func (*Trial) GetMetadata ¶
func (*Trial) GetRestarts ¶
func (*Trial) GetRunnerState ¶
func (*Trial) GetSearcherMetricValue ¶
func (*Trial) GetStartTime ¶
func (*Trial) GetSummaryMetrics ¶
func (*Trial) GetTaskIds ¶
func (*Trial) GetTotalBatchesProcessed ¶
func (*Trial) GetTotalCheckpointSize ¶
func (*Trial) GetWallClockTime ¶
func (*Trial) GetWarmStartCheckpointUuid ¶
func (*Trial) ProtoMessage ¶
func (*Trial) ProtoMessage()
func (*Trial) ProtoReflect ¶
func (x *Trial) ProtoReflect() protoreflect.Message
type TrialEarlyExit ¶
type TrialEarlyExit struct { // The reason for the exit. Reason TrialEarlyExit_ExitedReason `protobuf:"varint,1,opt,name=reason,proto3,enum=determined.trial.v1.TrialEarlyExit_ExitedReason" json:"reason,omitempty"` // contains filtered or unexported fields }
Signals to the experiment the trial early exited.
func (*TrialEarlyExit) Descriptor
deprecated
func (*TrialEarlyExit) Descriptor() ([]byte, []int)
Deprecated: Use TrialEarlyExit.ProtoReflect.Descriptor instead.
func (*TrialEarlyExit) GetReason ¶
func (x *TrialEarlyExit) GetReason() TrialEarlyExit_ExitedReason
func (*TrialEarlyExit) ProtoMessage ¶
func (*TrialEarlyExit) ProtoMessage()
func (*TrialEarlyExit) ProtoReflect ¶
func (x *TrialEarlyExit) ProtoReflect() protoreflect.Message
func (*TrialEarlyExit) Reset ¶
func (x *TrialEarlyExit) Reset()
func (*TrialEarlyExit) String ¶
func (x *TrialEarlyExit) String() string
type TrialEarlyExit_ExitedReason ¶
type TrialEarlyExit_ExitedReason int32
The reason for an early exit.
const ( // Zero-value (not allowed). TrialEarlyExit_EXITED_REASON_UNSPECIFIED TrialEarlyExit_ExitedReason = 0 // Indicates the trial exited due to an invalid hyperparameter. TrialEarlyExit_EXITED_REASON_INVALID_HP TrialEarlyExit_ExitedReason = 1 // Indicates the trial exited due to an invalid hyperparameter // in the trial init. TrialEarlyExit_EXITED_REASON_INIT_INVALID_HP TrialEarlyExit_ExitedReason = 3 )
func (TrialEarlyExit_ExitedReason) Descriptor ¶
func (TrialEarlyExit_ExitedReason) Descriptor() protoreflect.EnumDescriptor
func (TrialEarlyExit_ExitedReason) Enum ¶
func (x TrialEarlyExit_ExitedReason) Enum() *TrialEarlyExit_ExitedReason
func (TrialEarlyExit_ExitedReason) EnumDescriptor
deprecated
func (TrialEarlyExit_ExitedReason) EnumDescriptor() ([]byte, []int)
Deprecated: Use TrialEarlyExit_ExitedReason.Descriptor instead.
func (TrialEarlyExit_ExitedReason) Number ¶
func (x TrialEarlyExit_ExitedReason) Number() protoreflect.EnumNumber
func (TrialEarlyExit_ExitedReason) String ¶
func (x TrialEarlyExit_ExitedReason) String() string
func (TrialEarlyExit_ExitedReason) Type ¶
func (TrialEarlyExit_ExitedReason) Type() protoreflect.EnumType
type TrialMetrics ¶
type TrialMetrics struct { // The trial associated with these metrics. TrialId int32 `protobuf:"varint,1,opt,name=trial_id,json=trialId,proto3" json:"trial_id,omitempty"` // The trial run associated with these metrics. TrialRunId int32 `protobuf:"varint,2,opt,name=trial_run_id,json=trialRunId,proto3" json:"trial_run_id,omitempty"` // The number of batches trained on when these metrics were reported. StepsCompleted *int32 `protobuf:"varint,3,opt,name=steps_completed,json=stepsCompleted,proto3,oneof" json:"steps_completed,omitempty"` // The client-reported time associated with these metrics. ReportTime *timestamp.Timestamp `protobuf:"bytes,4,opt,name=report_time,json=reportTime,proto3,oneof" json:"report_time,omitempty"` // The metrics for this bit of training, including: // - avg_metrics: metrics reduced over the reporting period). // - batch_metrics: (optional) per-batch metrics. Metrics *commonv1.Metrics `protobuf:"bytes,9,opt,name=metrics,proto3" json:"metrics,omitempty"` // contains filtered or unexported fields }
Metrics from the trial some duration of training.
func (*TrialMetrics) Descriptor
deprecated
func (*TrialMetrics) Descriptor() ([]byte, []int)
Deprecated: Use TrialMetrics.ProtoReflect.Descriptor instead.
func (*TrialMetrics) GetMetrics ¶
func (x *TrialMetrics) GetMetrics() *commonv1.Metrics
func (*TrialMetrics) GetReportTime ¶
func (x *TrialMetrics) GetReportTime() *timestamp.Timestamp
func (*TrialMetrics) GetStepsCompleted ¶
func (x *TrialMetrics) GetStepsCompleted() int32
func (*TrialMetrics) GetTrialId ¶
func (x *TrialMetrics) GetTrialId() int32
func (*TrialMetrics) GetTrialRunId ¶
func (x *TrialMetrics) GetTrialRunId() int32
func (*TrialMetrics) ProtoMessage ¶
func (*TrialMetrics) ProtoMessage()
func (*TrialMetrics) ProtoReflect ¶
func (x *TrialMetrics) ProtoReflect() protoreflect.Message
func (*TrialMetrics) Reset ¶
func (x *TrialMetrics) Reset()
func (*TrialMetrics) String ¶
func (x *TrialMetrics) String() string
type TrialProfilerMetricLabels ¶
type TrialProfilerMetricLabels struct { // The ID of the trial. TrialId int32 `protobuf:"varint,1,opt,name=trial_id,json=trialId,proto3" json:"trial_id,omitempty"` // The name of the metric. Name string `protobuf:"bytes,2,opt,name=name,proto3" json:"name,omitempty"` // The agent ID associated with the metric. AgentId string `protobuf:"bytes,3,opt,name=agent_id,json=agentId,proto3" json:"agent_id,omitempty"` // The GPU UUID associated with the metric. GpuUuid string `protobuf:"bytes,4,opt,name=gpu_uuid,json=gpuUuid,proto3" json:"gpu_uuid,omitempty"` // The type of the metric. MetricType TrialProfilerMetricLabels_ProfilerMetricType `` /* 162-byte string literal not displayed */ // contains filtered or unexported fields }
TrialProfilerMetricLabels are the labels for a single series, where a series is a defined as all metrics sharing a distinct set of labels
func (*TrialProfilerMetricLabels) Descriptor
deprecated
func (*TrialProfilerMetricLabels) Descriptor() ([]byte, []int)
Deprecated: Use TrialProfilerMetricLabels.ProtoReflect.Descriptor instead.
func (*TrialProfilerMetricLabels) GetAgentId ¶
func (x *TrialProfilerMetricLabels) GetAgentId() string
func (*TrialProfilerMetricLabels) GetGpuUuid ¶
func (x *TrialProfilerMetricLabels) GetGpuUuid() string
func (*TrialProfilerMetricLabels) GetMetricType ¶
func (x *TrialProfilerMetricLabels) GetMetricType() TrialProfilerMetricLabels_ProfilerMetricType
func (*TrialProfilerMetricLabels) GetName ¶
func (x *TrialProfilerMetricLabels) GetName() string
func (*TrialProfilerMetricLabels) GetTrialId ¶
func (x *TrialProfilerMetricLabels) GetTrialId() int32
func (*TrialProfilerMetricLabels) ProtoMessage ¶
func (*TrialProfilerMetricLabels) ProtoMessage()
func (*TrialProfilerMetricLabels) ProtoReflect ¶
func (x *TrialProfilerMetricLabels) ProtoReflect() protoreflect.Message
func (*TrialProfilerMetricLabels) Reset ¶
func (x *TrialProfilerMetricLabels) Reset()
func (*TrialProfilerMetricLabels) String ¶
func (x *TrialProfilerMetricLabels) String() string
type TrialProfilerMetricLabels_ProfilerMetricType ¶
type TrialProfilerMetricLabels_ProfilerMetricType int32
To distinguish the 2 different categories of metrics.
const ( // Zero-value (not allowed). TrialProfilerMetricLabels_PROFILER_METRIC_TYPE_UNSPECIFIED TrialProfilerMetricLabels_ProfilerMetricType = 0 // For systems metrics, like GPU utilization or memory. TrialProfilerMetricLabels_PROFILER_METRIC_TYPE_SYSTEM TrialProfilerMetricLabels_ProfilerMetricType = 1 // For timing metrics, like how long a backwards pass or getting a batch // from the dataloader took. TrialProfilerMetricLabels_PROFILER_METRIC_TYPE_TIMING TrialProfilerMetricLabels_ProfilerMetricType = 2 // For other miscellaneous metrics. TrialProfilerMetricLabels_PROFILER_METRIC_TYPE_MISC TrialProfilerMetricLabels_ProfilerMetricType = 3 )
func (TrialProfilerMetricLabels_ProfilerMetricType) Descriptor ¶
func (TrialProfilerMetricLabels_ProfilerMetricType) Descriptor() protoreflect.EnumDescriptor
func (TrialProfilerMetricLabels_ProfilerMetricType) EnumDescriptor
deprecated
func (TrialProfilerMetricLabels_ProfilerMetricType) EnumDescriptor() ([]byte, []int)
Deprecated: Use TrialProfilerMetricLabels_ProfilerMetricType.Descriptor instead.
func (TrialProfilerMetricLabels_ProfilerMetricType) Number ¶
func (x TrialProfilerMetricLabels_ProfilerMetricType) Number() protoreflect.EnumNumber
func (TrialProfilerMetricLabels_ProfilerMetricType) String ¶
func (x TrialProfilerMetricLabels_ProfilerMetricType) String() string
func (TrialProfilerMetricLabels_ProfilerMetricType) Type ¶
func (TrialProfilerMetricLabels_ProfilerMetricType) Type() protoreflect.EnumType
type TrialProfilerMetricsBatch ¶
type TrialProfilerMetricsBatch struct { // The measurement for a reading, repeated for the batch of metrics. Values []float32 `protobuf:"fixed32,1,rep,packed,name=values,proto3" json:"values,omitempty"` // The batch at which a reading occurred, repeated for the batch of metrics. Batches []int32 `protobuf:"varint,2,rep,packed,name=batches,proto3" json:"batches,omitempty"` // The timestamp at which a reading occurred, repeated for the batch of // metrics. Timestamps []*timestamp.Timestamp `protobuf:"bytes,3,rep,name=timestamps,proto3" json:"timestamps,omitempty"` // The labels for this series. Labels *TrialProfilerMetricLabels `protobuf:"bytes,4,opt,name=labels,proto3" json:"labels,omitempty"` // contains filtered or unexported fields }
TrialProfilerMetricsBatch is a batch of trial profiler metrics. A batch will contain metrics pertaining to a single series. The fields values, batches and timestamps will be equal length arrays with each index corresponding to a reading.
func (*TrialProfilerMetricsBatch) Descriptor
deprecated
func (*TrialProfilerMetricsBatch) Descriptor() ([]byte, []int)
Deprecated: Use TrialProfilerMetricsBatch.ProtoReflect.Descriptor instead.
func (*TrialProfilerMetricsBatch) GetBatches ¶
func (x *TrialProfilerMetricsBatch) GetBatches() []int32
func (*TrialProfilerMetricsBatch) GetLabels ¶
func (x *TrialProfilerMetricsBatch) GetLabels() *TrialProfilerMetricLabels
func (*TrialProfilerMetricsBatch) GetTimestamps ¶
func (x *TrialProfilerMetricsBatch) GetTimestamps() []*timestamp.Timestamp
func (*TrialProfilerMetricsBatch) GetValues ¶
func (x *TrialProfilerMetricsBatch) GetValues() []float32
func (*TrialProfilerMetricsBatch) ProtoMessage ¶
func (*TrialProfilerMetricsBatch) ProtoMessage()
func (*TrialProfilerMetricsBatch) ProtoReflect ¶
func (x *TrialProfilerMetricsBatch) ProtoReflect() protoreflect.Message
func (*TrialProfilerMetricsBatch) Reset ¶
func (x *TrialProfilerMetricsBatch) Reset()
func (*TrialProfilerMetricsBatch) String ¶
func (x *TrialProfilerMetricsBatch) String() string
type TrialRunnerMetadata ¶
type TrialRunnerMetadata struct { // The state of the trial runner. State string `protobuf:"bytes,1,opt,name=state,proto3" json:"state,omitempty"` // contains filtered or unexported fields }
The metadata pertaining to the current running task for a trial.
func (*TrialRunnerMetadata) Descriptor
deprecated
func (*TrialRunnerMetadata) Descriptor() ([]byte, []int)
Deprecated: Use TrialRunnerMetadata.ProtoReflect.Descriptor instead.
func (*TrialRunnerMetadata) GetState ¶
func (x *TrialRunnerMetadata) GetState() string
func (*TrialRunnerMetadata) ProtoMessage ¶
func (*TrialRunnerMetadata) ProtoMessage()
func (*TrialRunnerMetadata) ProtoReflect ¶
func (x *TrialRunnerMetadata) ProtoReflect() protoreflect.Message
func (*TrialRunnerMetadata) Reset ¶
func (x *TrialRunnerMetadata) Reset()
func (*TrialRunnerMetadata) String ¶
func (x *TrialRunnerMetadata) String() string
type TrialSourceInfo ¶
type TrialSourceInfo struct { // ID of the trial. TrialId int32 `protobuf:"varint,1,opt,name=trial_id,json=trialId,proto3" json:"trial_id,omitempty"` // UUID of the checkpoint. CheckpointUuid string `protobuf:"bytes,2,opt,name=checkpoint_uuid,json=checkpointUuid,proto3" json:"checkpoint_uuid,omitempty"` // Source `id“ for the model which generated the checkpoint (if applicable) ModelId *int32 `protobuf:"varint,3,opt,name=model_id,json=modelId,proto3,oneof" json:"model_id,omitempty"` // Source `version` in the model_version version field which generated the // checkpoint (if applicable) ModelVersion *int32 `protobuf:"varint,4,opt,name=model_version,json=modelVersion,proto3,oneof" json:"model_version,omitempty"` // Type for this trial_source_info TrialSourceInfoType TrialSourceInfoType `` /* 168-byte string literal not displayed */ // contains filtered or unexported fields }
Denotes a connection between a given trial and a checkpoint or model_version
func (*TrialSourceInfo) Descriptor
deprecated
func (*TrialSourceInfo) Descriptor() ([]byte, []int)
Deprecated: Use TrialSourceInfo.ProtoReflect.Descriptor instead.
func (*TrialSourceInfo) GetCheckpointUuid ¶
func (x *TrialSourceInfo) GetCheckpointUuid() string
func (*TrialSourceInfo) GetModelId ¶
func (x *TrialSourceInfo) GetModelId() int32
func (*TrialSourceInfo) GetModelVersion ¶
func (x *TrialSourceInfo) GetModelVersion() int32
func (*TrialSourceInfo) GetTrialId ¶
func (x *TrialSourceInfo) GetTrialId() int32
func (*TrialSourceInfo) GetTrialSourceInfoType ¶
func (x *TrialSourceInfo) GetTrialSourceInfoType() TrialSourceInfoType
func (*TrialSourceInfo) ProtoMessage ¶
func (*TrialSourceInfo) ProtoMessage()
func (*TrialSourceInfo) ProtoReflect ¶
func (x *TrialSourceInfo) ProtoReflect() protoreflect.Message
func (*TrialSourceInfo) Reset ¶
func (x *TrialSourceInfo) Reset()
func (*TrialSourceInfo) String ¶
func (x *TrialSourceInfo) String() string
type TrialSourceInfoType ¶
type TrialSourceInfoType int32
TrialSourceInfoType is the type of the TrialSourceInfo, which serves as a link between a trial and a checkpoint or model version
const ( // The type is unspecified TrialSourceInfoType_TRIAL_SOURCE_INFO_TYPE_UNSPECIFIED TrialSourceInfoType = 0 // "Inference" Trial Source Info Type, used for batch inference TrialSourceInfoType_TRIAL_SOURCE_INFO_TYPE_INFERENCE TrialSourceInfoType = 1 // "Fine Tuning" Trial Source Info Type, used in model hub TrialSourceInfoType_TRIAL_SOURCE_INFO_TYPE_FINE_TUNING TrialSourceInfoType = 2 )
func (TrialSourceInfoType) Descriptor ¶
func (TrialSourceInfoType) Descriptor() protoreflect.EnumDescriptor
func (TrialSourceInfoType) Enum ¶
func (x TrialSourceInfoType) Enum() *TrialSourceInfoType
func (TrialSourceInfoType) EnumDescriptor
deprecated
func (TrialSourceInfoType) EnumDescriptor() ([]byte, []int)
Deprecated: Use TrialSourceInfoType.Descriptor instead.
func (TrialSourceInfoType) Number ¶
func (x TrialSourceInfoType) Number() protoreflect.EnumNumber
func (TrialSourceInfoType) String ¶
func (x TrialSourceInfoType) String() string
func (TrialSourceInfoType) Type ¶
func (TrialSourceInfoType) Type() protoreflect.EnumType