pub struct Run {Show 18 fields
pub run_id: String,
pub status: RunStatus,
pub error: Option<RunError>,
pub skip_reason: Option<String>,
pub started_at: Option<DateTime<Utc>>,
pub ended_at: Option<DateTime<Utc>>,
pub model: Option<Model>,
pub task: Option<Task>,
pub harness: Option<Harness>,
pub generation: Option<Generation>,
pub trial: Option<Trial>,
pub env: Option<Env>,
pub calls: Option<String>,
pub artifacts: Vec<String>,
pub attachments: Vec<Attachment>,
pub metrics: BTreeMap<String, f64>,
pub ext: Ext,
pub meta: Option<Value>,
}Expand description
One run of an Eval: one execution, its conditions, its files and what
was measured directly from it. A run records what happened, never
whether it passed; the verdict is a Card’s run_results.
Fields§
§run_id: StringRun identifier, unique within the Eval record. One URL segment:
non-empty, no /, at most 200 bytes.
status: RunStatusThe state of the execution: ok, error or skipped.
error: Option<RunError>How the run failed. Present exactly when status is error.
skip_reason: Option<String>Why the producer did not execute the run. Present exactly when
status is skipped.
started_at: Option<DateTime<Utc>>When the run started.
ended_at: Option<DateTime<Utc>>When the run ended.
model: Option<Model>The model this run used. Filled from the Eval header when omitted.
task: Option<Task>The task this run attempted. Filled from the Eval header when omitted.
harness: Option<Harness>The harness that drove this run. Filled from the Eval header when omitted.
generation: Option<Generation>The sampling parameters of this run. Filled from the Eval header when omitted.
trial: Option<Trial>How this run’s attempt was made. Filled from the Eval header when omitted.
env: Option<Env>The environment this run executed in. Filled from the Eval header when omitted.
calls: Option<String>This run’s attachments[].path holding the model calls made during the run.
artifacts: Vec<String>This run’s attachments[].path entries of artifacts the run produced.
attachments: Vec<Attachment>Files this run refers to. calls, artifacts[] and error.log point here.
metrics: BTreeMap<String, f64>Values measured directly from the run, keyed by metric registry id
{ns}/{name} (for example core/tokens_out). Grader-dependent
values belong to a Card’s run_results, not here.
ext: ExtProducer-private extensions keyed by namespace.
meta: Option<Value>Free-form JSON kept with the run and returned verbatim. The hub never validates, indexes or queries it.
Trait Implementations§
Source§impl<'de> Deserialize<'de> for Run
impl<'de> Deserialize<'de> for Run
Source§fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
Source§impl JsonSchema for Run
impl JsonSchema for Run
Source§fn schema_id() -> Cow<'static, str>
fn schema_id() -> Cow<'static, str>
Source§fn json_schema(generator: &mut SchemaGenerator) -> Schema
fn json_schema(generator: &mut SchemaGenerator) -> Schema
Source§fn inline_schema() -> bool
fn inline_schema() -> bool
$ref keyword. Read more