package brain

Get desktop application:
View/edit binary Protocol Buffers messages

• rpc persist_metrics (JobMetrics, google.protobuf.Empty)

  brain.proto:197

  message JobMetrics

  brain.proto:92

  • string data_store = 1

  • optional JobMeta job_meta = 2

  • MetricsType metrics_type = 3

  • oneof metrics

    • TrainingHyperParams training_hyper_params = 4

    • WorkflowFeature workflow_feature = 5

    • TrainingSetFeature training_set_feature = 6

    • ModelFeature model_feature = 7

    • RuntimeInfo runtime_info = 8

    • string job_exit_reason = 9

    • string extra_info = 10

    • string type = 11

    • string resource = 12

    • string customized_data = 13

    • JobOptimization job_optimization = 14

• rpc optimize (OptimizeRequest, OptimizeResponse)

  brain.proto:198

  message OptimizeRequest

  brain.proto:142

  • string type = 1

  • optional OptimizeConfig config = 2

  • repeated OptimizeJobMeta jobs = 3

  message OptimizeResponse

  brain.proto:184

  • optional Response response = 1

  • repeated JobOptimizePlan job_optimize_plans = 2

• rpc get_job_metrics (JobMetricsRequest, JobMetricsResponse)

  brain.proto:199

  message JobMetricsRequest

  brain.proto:189

  • string job_uuid = 1

  message JobMetricsResponse

  brain.proto:191

  • optional Response response = 1

  • string job_metrics = 2

Used in: JobMetrics

• string uuid = 1

  The uuid of k8s training job

• string name = 2

  The name of k8s training job

• string user = 3

  The user id of k8s training job

• string cluster = 4

  The cluster of k8s training job

• string namespace = 5

  The namespace of k8s training job

Used in: JobMetrics

• int64 timestamp = 1

• optional OptimizeConfig config = 2

• optional JobOptimizePlan plan = 3

• repeated JobState job_states = 4

Used in: JobOptimization, OptimizeResponse

• uint64 timestamp = 1

• optional JobResource resource = 2

• optional OptimizeJobMeta job = 3

Used in: JobOptimizePlan

• map<string, TaskGroupResource> task_group_resources = 1

• map<string, PodResource> pod_resources = 2

Used in: JobOptimization, OptimizeJobMeta

• int64 timestamp = 1

• map<string, PodState> pods = 2

• float speed = 3

• map<string, string> customized_data = 4

Used in: JobMetrics

• Training_Hyper_Params = 0

• Workflow_Feature = 1

• Training_Set_Feature = 2

• Model_Feature = 3

• Runtime_Info = 4

• Job_Exit_Reason = 5

• Optimization_Result = 6

• Type = 7

• Resource = 8

• Customized_Data = 9

• Optimization = 10

Used in: JobMetrics

• int64 variable_count = 1

• int64 op_count = 2

• int64 embedding_dimension = 3

• int64 total_variable_size = 4

• int64 max_variable_size = 5

• int64 update_op_count = 6

• int64 read_op_count = 7

• int64 input_fetch_dur = 8

• int64 flops = 9

• int64 recv_op_count = 10

• repeated int64 kv_embedding_dims = 11

• map<string, int64> tensor_alloc_bytes = 12

Used in: JobOptimization, OptimizeRequest

• string optimizer_config_retriever = 1

• string data_store = 2

• string brain_processor = 3

• map<string, string> customized_config = 4

Used in: JobOptimizePlan, OptimizeRequest

• string uid = 1

• string cluster = 2

• string namespace = 3

• optional JobState state = 4

Used in: RuntimeInfo

• string pod_name = 1

• string pod_ip = 2

• string node_ip = 3

• string host_name = 4

• string namespace = 5

• bool is_mixed = 6

• double mem_usage = 7

• double cpu_usage = 8

Used in: JobResource, PodState, TaskGroupResource

• int64 memory = 1

  unit MB

• float cpu = 2

  The number of cores

• float gpu = 3

• string gpu_type = 4

• string priority = 5

Used in: JobState

• string name = 1

• string uid = 2

• string type = 3

• bool is_oom = 4

• optional PodResource resource = 5

• optional PodResource used_resource = 6

• map<string, string> customized_data = 7

Used in: JobMetricsResponse, OptimizeResponse

• bool success = 1

• string reason = 2

Used in: JobMetrics

• int64 global_step = 1

• repeated PodMeta running_pods = 2

• int64 time_stamp = 3

• float speed = 4

Used in: JobResource

• int64 count = 1

• optional PodResource resource = 2

Used in: JobMetrics

• int64 batch_size = 1

• int64 epoch = 2

• int64 max_steps = 3

Used in: JobMetrics

• int64 dataset_size = 1

• string dataset_name = 2

• int64 sparse_item_count = 3

• string sparse_features = 4

• string sparse_feature_groups = 5

• string sparse_feature_shapes = 6

• string dense_features = 7

• string dense_feature_shapes = 8

• int64 storage_size = 9

Used in: JobMetrics

• string job_name = 1

• string user_id = 2

• string code_address = 3

• string workflow_id = 4

• string node_id = 5

• string odps_project = 6

• bool is_prod = 7