Struct JobDefinition 

pub struct JobDefinition {

    pub name: String,
    pub description: Option<String>,
    pub flow_type: FlowType,
    pub env_name: Option<String>,
    pub task_name: Option<String>,
    pub algorithm: Option<String>,
    pub policy: Option<String>,
    pub seed: Option<i64>,
    pub environment_map: Option<String>,
    pub total_timesteps: Option<i64>,
    pub learning_rate: Option<f64>,
    pub linear_lr_schedule: Option<bool>,
    pub n_steps: Option<i64>,
    pub batch_size: Option<i64>,
    pub n_epochs: Option<i64>,
    pub gamma: Option<f64>,
    pub gae_lambda: Option<f64>,
    pub clip_range: Option<f64>,
    pub ent_coef: Option<f64>,
    pub vf_coef: Option<f64>,
    pub max_grad_norm: Option<f64>,
    pub patience: Option<i64>,
    pub min_delta: Option<f64>,
    pub n_envs: Option<i64>,
    pub n_trials: Option<i64>,
    pub tune_timesteps: Option<i64>,
    pub eval_eps: Option<i64>,
    pub search_space: Option<Vec<SearchSpaceEntry>>,
    pub source_path: PathBuf,
}

A training job definition, loaded from configs/rl/jobs/<name>.yml.

Fields

name: String

description: Option<String>

flow_type: FlowType

env_name: Option<String>

Name of the simulation environment to train in. If omitted, it’s resolved from the selected task.

task_name: Option<String>

Name of the task/curriculum to train on, selected from the project’s task catalog.

algorithm: Option<String>

Name of the reinforcement learning algorithm to use for this job.

policy: Option<String>

Name of the policy network architecture to use. If omitted, it’s auto-detected from the environment.

seed: Option<i64>

environment_map: Option<String>

total_timesteps: Option<i64>

learning_rate: Option<f64>

linear_lr_schedule: Option<bool>

n_steps: Option<i64>

batch_size: Option<i64>

n_epochs: Option<i64>

gamma: Option<f64>

gae_lambda: Option<f64>

clip_range: Option<f64>

ent_coef: Option<f64>

vf_coef: Option<f64>

max_grad_norm: Option<f64>

patience: Option<i64>

Number of evaluations with no improvement to wait before stopping training early.

min_delta: Option<f64>

n_envs: Option<i64>

Number of parallel environments to run during training.

n_trials: Option<i64>

tune_timesteps: Option<i64>

eval_eps: Option<i64>

search_space: Option<Vec<SearchSpaceEntry>>

Hyperparameter search space to explore for a tuning job. If omitted, a default search space is used.

source_path: PathBuf

Implementations

impl JobDefinition

pub fn validate( &self, project_root: &Path, resume_run_id: Option<&str>, ) -> Result<()>

Validate this job against mecha10-job-validation’s flow_type requirement table and its by-algorithm hyperparameter requirement table, plus a task_name-exists check against the project’s own task catalog. Algorithm-awareness (e.g. ppo needing n_steps, act needing none of these fields) lives entirely in mecha10-job-validation‘s AlgorithmField table - this method only composes the two generic validators’ results, the same way compose_node_schema composes a base schema with a node’s own custom schema rather than validating them separately.

pub fn resolve_env_name(&self, project_root: &Path) -> Option<String>

Resolve which mecha10_rl.envs registry name this job targets: an explicit env_name wins, else it’s read from task_name’s own task.json. None when neither is available.

pub fn supports_remote(&self) -> bool

Whether this job can be dispatched remotely with today’s CreateJobRequest/create_job contract. tune-then-train/train-from-tune aren’t dispatchable yet and stay local-only.

pub fn to_create_job_request( &self, project_root: &Path, robot_id: &str, project_name: &str, resume_run_id: Option<&str>, ) -> Result<CreateJobRequest>

Build the remote dispatch request body. Only meaningful when supports_remote() is true; callers should check that first.

Trait Implementations

impl Clone for JobDefinition

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for JobDefinition

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for JobDefinition

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl JsonSchema for JobDefinition

fn schema_name() -> Cow<'static, str>

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn inline_schema() -> bool

Whether JSON Schemas generated for this type should be included directly in parent schemas, rather than being re-used where possible using the $ref keyword.