Do not produce ProgressionPlot if all steps are NaN (#4769)

Summary:
Pull Request resolved: #4769

Prevents confusing plots like this one:
{F1984569487}

Reviewed By: mpolson64

Differential Revision: D90715040

fbshipit-source-id: cb34c04dd74fddf29620411d24593b1e90900b6e

Author: saitcakmak

ax/analysis/plotly/progression.py

@@ -75,7 +75,7 @@ def validate_applicable_state(
     ) -> str | None:
         """
         ProgressionPlot requires an Experiment with data that has a "step"
-        column.
+        column with valid (non-NaN) values.
         """
         if (
             experiment_invalid_reason := validate_experiment(
@@ -90,6 +90,20 @@ def validate_applicable_state(
         if not data.has_step_column:
             return "Requires data to have a column 'step.'"
 
+        # Check if the step column has any valid (non-NaN) values
+        metric_name = self._metric_name or select_metric(
+            experiment=none_throws(experiment)
+        )
+        df = none_throws(experiment).lookup_data().full_df
+        metric_df = df[df["metric_name"] == metric_name]
+
+        if metric_df.empty:
+            return f"No data found for metric '{metric_name}'."
+
+        # Check if all step values are NaN
+        if metric_df[MAP_KEY].isna().all():
+            return f"All progression values for metric '{metric_name}' are NaN."
+
     @override
     def compute(
         self,

ax/analysis/plotly/tests/test_progression.py

@@ -5,11 +5,13 @@
 
 # pyre-strict
 
+import numpy as np
 import pandas as pd
 from ax.analysis.plotly.progression import (
     _calculate_wallclock_timeseries,
     ProgressionPlot,
 )
+from ax.core.data import Data, MAP_KEY
 from ax.utils.common.testutils import TestCase
 from ax.utils.testing.core_stubs import (
     get_branin_experiment,
@@ -35,6 +37,28 @@ def test_validate_applicable_state(self) -> None:
             state = plot.validate_applicable_state(experiment=experiment)
             self.assertEqual(state, "Requires data to have a column 'step.'")
 
+        with self.subTest("All step values are NaN"):
+            # Create a new experiment with map data where all MAP_KEY values are NaN
+            experiment = get_test_map_data_experiment(
+                num_trials=2, num_fetches=3, num_complete=2
+            )
+
+            # Replace all MAP_KEY values with NaN in the fetched data
+            original_data = experiment.fetch_data()
+            modified_df = original_data.full_df.copy()
+            modified_df[MAP_KEY] = np.nan
+
+            # Create a new Data object and attach it
+            nan_data = Data(df=modified_df)
+            experiment.data = nan_data
+
+            # Validate that progression plot is not applicable
+            plot = ProgressionPlot(metric_name="branin_map")
+            state = plot.validate_applicable_state(experiment=experiment)
+            self.assertEqual(
+                state, "All progression values for metric 'branin_map' are NaN."
+            )
+
     def test_compute(self) -> None:
         analysis = ProgressionPlot(metric_name="branin_map")
 

ax/analysis/results.py

@@ -29,6 +29,7 @@
 from ax.core.analysis_card import AnalysisCardGroup
 from ax.core.arm import Arm
 from ax.core.batch_trial import BatchTrial
+from ax.core.data import MAP_KEY
 from ax.core.experiment import Experiment
 from ax.core.map_metric import MapMetric
 from ax.core.outcome_constraint import ScalarizedOutcomeConstraint
@@ -246,12 +247,16 @@ def compute(
             adapter=adapter,
         )
 
-        # Compute progression plots for MapMetrics (learning curves)
+        # Compute progression plots if there is curve data.
         progression_group = None
         data = experiment.lookup_data()
         metrics = experiment.metrics.values()
         map_metrics = [m for m in metrics if isinstance(m, MapMetric)]
-        if data.has_step_column and len(map_metrics) > 0:
+        if (
+            data.has_step_column
+            and data.full_df[MAP_KEY].notna().any()
+            and len(map_metrics) > 0
+        ):
             progression_cards = [
                 ProgressionPlot(
                     metric_name=m.name, by_wallclock_time=by_wallclock_time