remove step from trial

alphatrion/log/log.py

@@ -105,7 +105,6 @@ async def log_metrics(metrics: dict[str, float]):
     should_checkpoint = False
     should_early_stop = False
     should_stop_on_target = False
-    step = trial.increment_step()
     for key, value in metrics.items():
         runtime._metadb.create_metric(
             key=key,
@@ -114,7 +113,6 @@ async def log_metrics(metrics: dict[str, float]):
             experiment_id=exp.id,
             trial_id=trial_id,
             run_id=run_id,
-            step=step,
         )
 
         # TODO: should we save the checkpoint path for the best metric?
@@ -128,6 +126,7 @@ async def log_metrics(metrics: dict[str, float]):
             metric_key=key, metric_value=value
         )
 
+    # TODO: we should save the artifact path in the metrics as well.
     if should_checkpoint:
         await log_artifact(
             paths=trial.config().checkpoint.path,

alphatrion/metadata/base.py

@@ -111,7 +111,6 @@ def create_metric(
         run_id: uuid.UUID,
         key: str,
         value: float,
-        step: int | None = None,
     ) -> int:
         raise NotImplementedError("Subclasses must implement this method.")
 

alphatrion/metadata/sql.py

@@ -370,7 +370,6 @@ def create_metric(
         run_id: uuid.UUID,
         key: str,
         value: float,
-        step: int,
     ) -> uuid.UUID:
         session = self._session()
         new_metric = Metric(
@@ -380,7 +379,6 @@ def create_metric(
             run_id=run_id,
             key=key,
             value=value,
-            step=step,
         )
         session.add(new_metric)
         session.commit()

alphatrion/metadata/sql_models.py

@@ -151,6 +151,7 @@ class Model(Base):
     is_del = Column(Integer, default=0, comment="0 for not deleted, 1 for deleted")
 
 
+# TODO: key, project_id, experiment_id, trial_id, run_id should be unique together
 class Metric(Base):
     __tablename__ = "metrics"
 
@@ -161,5 +162,4 @@ class Metric(Base):
     experiment_id = Column(UUID(as_uuid=True), nullable=False)
     trial_id = Column(UUID(as_uuid=True), nullable=False)
     run_id = Column(UUID(as_uuid=True), nullable=False)
-    step = Column(Integer, nullable=False, default=0)
     created_at = Column(DateTime(timezone=True), default=datetime.now(UTC))

alphatrion/server/graphql/resolvers.py

@@ -183,7 +183,6 @@ def list_trial_metrics(
                 experiment_id=m.experiment_id,
                 trial_id=m.trial_id,
                 run_id=m.run_id,
-                step=m.step,
                 created_at=m.created_at,
             )
             for m in metrics

alphatrion/server/graphql/types.py

@@ -82,5 +82,4 @@ class Metric:
     experiment_id: strawberry.ID
     trial_id: strawberry.ID
     run_id: strawberry.ID
-    step: int
     created_at: datetime

alphatrion/trial/trial.py

@@ -122,8 +122,6 @@ class Trial:
         "_exp_id",
         "_config",
         "_runtime",
-        # step is used to track the round, e.g. the step in metric logging.
-        "_step",
         "_context",
         "_token",
         # _meta stores the runtime meta information of the trial.
@@ -148,7 +146,6 @@ def __init__(self, exp_id: int, config: TrialConfig | None = None):
         self._exp_id = exp_id
         self._config = config or TrialConfig()
         self._runtime = global_runtime()
-        self._step = 0
         self._construct_meta()
         self._runs = dict[uuid.UUID, Run]()
         self._early_stopping_counter = 0
@@ -350,10 +347,6 @@ def _stop(self):
     def _get_obj(self):
         return self._runtime.metadb.get_trial(trial_id=self._id)
 
-    def increment_step(self) -> int:
-        self._step += 1
-        return self._step
-
     def start_run(self, call_func: callable) -> Run:
         """Start a new run for the trial.
         :param call_func: a callable function that returns a coroutine.

dashboard/src/components/trials/trial-detail.tsx

@@ -38,7 +38,6 @@ function MetricsChart({ metrics }: { metrics: Metric[] }) {
         metricsByKey[key].push(m);
     });
 
-    Object.values(metricsByKey).forEach((arr) => arr.sort((a, b) => a.step - b.step));
     const keys = Object.keys(metricsByKey);
 
     return (

dashboard/src/graphql/queries.ts

@@ -139,7 +139,6 @@ export const LIST_TRIAL_METRICS = `
       experimentId
       trialId
       runId
-      step
       createdAt
     }
   }

dashboard/src/types/index.ts

@@ -58,7 +58,6 @@ export interface Metric {
     experimentId: string;
     trialId: string;
     runId: string;
-    step: number;
     createdAt: string;
 }
 

hack/seed.py

@@ -109,7 +109,6 @@ def generate_metric(runs: list[Run]) -> Metric:
         run_id=run.uuid,
         key=random.choice(["accuracy", "loss", "precision", "fitness"]),
         value=random.uniform(0, 1),
-        step=random.randint(1, 1000),
     )
 
 

migrations/versions/03410247c6b7_remove_step_from_trial.py

@@ -0,0 +1,32 @@
+"""remove step from trial
+
+Revision ID: 03410247c6b7
+Revises: c89fc8504699
+Create Date: 2026-01-16 15:17:18.057002
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '03410247c6b7'
+down_revision: Union[str, Sequence[str], None] = 'c89fc8504699'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column('metrics', 'step')
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column('metrics', sa.Column('step', sa.INTEGER(), autoincrement=False, nullable=False))
+    # ### end Alembic commands ###

tests/integration/server/test_graphql_query.py

@@ -304,7 +304,6 @@ def test_query_trial_metrics():
         run_id=uuid.uuid4(),
         key="accuracy",
         value=0.95,
-        step=0,
     )
     _ = metadb.create_metric(
         project_id=project_id,
@@ -313,7 +312,6 @@ def test_query_trial_metrics():
         run_id=uuid.uuid4(),
         key="accuracy",
         value=0.95,
-        step=1,
     )
     query = f"""
     query {{
@@ -325,7 +323,6 @@ def test_query_trial_metrics():
             experimentId
             trialId
             runId
-            step
             createdAt
         }}
     }}

tests/integration/test_log.py

@@ -115,10 +115,8 @@ async def log_metric(metrics: dict):
         assert len(metrics) == 2
         assert metrics[0].key == "accuracy"
         assert metrics[0].value == 0.95
-        assert metrics[0].step == 1
         assert metrics[1].key == "loss"
         assert metrics[1].value == 0.1
-        assert metrics[1].step == 1
         run_id_1 = metrics[0].run_id
         assert run_id_1 is not None
         assert metrics[0].run_id == metrics[1].run_id
@@ -130,7 +128,6 @@ async def log_metric(metrics: dict):
         assert len(metrics) == 3
         assert metrics[2].key == "accuracy"
         assert metrics[2].value == 0.96
-        assert metrics[2].step == 2
         run_id_2 = metrics[2].run_id
         assert run_id_2 is not None
         assert run_id_2 != run_id_1

tests/unit/metadata/test_sql.py

@@ -98,8 +98,8 @@ def test_create_metric(db):
     run_id = db.create_run(
         trial_id=trial_id, project_id=project_id, experiment_id=exp_id
     )
-    db.create_metric(project_id, exp_id, trial_id, run_id, "accuracy", 0.95, 1)
-    db.create_metric(project_id, exp_id, trial_id, run_id, "accuracy", 0.85, 2)
+    db.create_metric(project_id, exp_id, trial_id, run_id, "accuracy", 0.95)
+    db.create_metric(project_id, exp_id, trial_id, run_id, "accuracy", 0.85)
 
     metrics = db.list_metrics_by_trial_id(trial_id)
     assert len(metrics) == 2