Fix bug when given -inf or nan values

optuna_dashboard/_serializer.py

@@ -1,11 +1,11 @@
 import json
-import math
 from typing import Any
 from typing import Dict
 from typing import List
 from typing import Tuple
 from typing import Union
 
+import numpy as np
 from optuna.distributions import BaseDistribution
 from optuna.study import StudySummary
 from optuna.trial import FrozenTrial
@@ -14,8 +14,10 @@
 
 
 try:
+    from typing import Literal
     from typing import TypedDict
 except ImportError:
+    from typing_extensions import Literal  # type: ignore
     from typing_extensions import TypedDict
 
 
@@ -31,14 +33,7 @@
     "IntermediateValue",
     {
         "step": int,
-        "value": Union[float, str],
-    },
-)
-TrialParam = TypedDict(
-    "TrialParam",
-    {
-        "name": str,
-        "value": str,
+        "value": Union[float, Literal["inf", "-inf", "nan"]],
     },
 )
 
@@ -56,17 +51,6 @@ def serialize_attrs(attrs: Dict[str, Any]) -> List[Attribute]:
     return serialized
 
 
-def serialize_intermediate_values(values: Dict[int, float]) -> List[IntermediateValue]:
-    return [
-        {"step": step, "value": "inf" if math.isinf(value) else value}
-        for step, value in values.items()
-    ]
-
-
-def serialize_trial_params(params: Dict[str, Any]) -> List[TrialParam]:
-    return [{"name": name, "value": str(value)} for name, value in params.items()]
-
-
 def serialize_study_summary(summary: StudySummary) -> Dict[str, Any]:
     serialized = {
         "study_id": summary._study_id,
@@ -118,14 +102,39 @@ def serialize_frozen_trial(study_id: int, trial: FrozenTrial) -> Dict[str, Any]:
         "study_id": study_id,
         "number": trial.number,
         "state": trial.state.name.capitalize(),
-        "intermediate_values": serialize_intermediate_values(trial.intermediate_values),
-        "params": serialize_trial_params(trial.params),
+        "params": [
+            {"name": name, "value": str(value)} for name, value in trial.params.items()
+        ],
         "user_attrs": serialize_attrs(trial.user_attrs),
         "system_attrs": serialize_attrs(trial.system_attrs),
     }
 
+    serialized_intermediate_values: List[IntermediateValue] = []
+    for step, value in trial.intermediate_values.items():
+        serialized_value: Union[float, Literal["nan", "inf", "-inf"]]
+        if np.isnan(value):
+            serialized_value = "nan"
+        elif np.isposinf(value):
+            serialized_value = "inf"
+        elif np.isneginf(value):
+            serialized_value = "-inf"
+        else:
+            assert np.isfinite(value)
+            serialized_value = value
+        serialized_intermediate_values.append({"step": step, "value": serialized_value})
+    serialized["intermediate_values"] = serialized_intermediate_values
+
     if trial.values is not None:
-        serialized["values"] = ["inf" if math.isinf(v) else v for v in trial.values]
+        serialized_values: List[Union[float, Literal["inf", "-inf"]]] = []
+        for v in trial.values:
+            assert not np.isnan(v), "Should not detect nan value"
+            if np.isposinf(v):
+                serialized_values.append("inf")
+            elif np.isneginf(v):
+                serialized_values.append("-inf")
+            else:
+                serialized_values.append(v)
+        serialized["values"] = serialized_values
 
     if trial.datetime_start is not None:
         serialized["datetime_start"] = trial.datetime_start.isoformat()

optuna_dashboard/ts/apiClient.ts

@@ -7,7 +7,7 @@ interface TrialResponse {
   study_id: number
   number: number
   state: TrialState
-  values?: (number | "inf")[]
+  values?: TrialValueNumber[]
   intermediate_values: TrialIntermediateValue[]
   datetime_start?: string
   datetime_complete?: string

optuna_dashboard/ts/components/GraphEdf.tsx

@@ -66,7 +66,8 @@ const filterFunc = (trial: Trial, objectiveId: number): boolean => {
   return (
     trial.state === "Complete" &&
     trial.values !== undefined &&
-    trial.values[objectiveId] !== "inf"
+    trial.values[objectiveId] !== "inf" &&
+    trial.values[objectiveId] !== "-inf"
   )
 }
 

optuna_dashboard/ts/components/GraphHistory.tsx

@@ -180,7 +180,9 @@ const filterFunc = (trial: Trial, objectiveId: number): boolean => {
     return false
   }
   return (
-    trial.values.length > objectiveId && trial.values[objectiveId] !== "inf"
+    trial.values.length > objectiveId &&
+    trial.values[objectiveId] !== "inf" &&
+    trial.values[objectiveId] !== "-inf"
   )
 }
 

optuna_dashboard/ts/components/GraphIntermediateValues.tsx

@@ -58,7 +58,9 @@ const plotIntermediateValue = (trials: Trial[], mode: string) => {
       t.state == "Running"
   )
   const plotData: Partial<plotly.PlotData>[] = filteredTrials.map((trial) => {
-    const values = trial.intermediate_values.filter((iv) => iv.value !== "inf")
+    const values = trial.intermediate_values.filter(
+      (iv) => iv.value !== "inf" && iv.value !== "-inf" && iv.value !== "nan"
+    )
     return {
       x: values.map((iv) => iv.step),
       y: values.map((iv) => iv.value),

optuna_dashboard/ts/components/GraphParallelCoordinate.tsx

@@ -71,7 +71,9 @@ const filterFunc = (trial: Trial, objectiveId: number): boolean => {
     return false
   }
   return (
-    trial.values.length > objectiveId && trial.values[objectiveId] !== "inf"
+    trial.values.length > objectiveId &&
+    trial.values[objectiveId] !== "inf" &&
+    trial.values[objectiveId] !== "-inf"
   )
 }
 

optuna_dashboard/ts/components/GraphParetoFront.tsx

@@ -90,7 +90,7 @@ const filterFunc = (trial: Trial, directions: StudyDirection[]): boolean => {
     trial.state === "Complete" &&
     trial.values !== undefined &&
     trial.values.length === directions.length &&
-    trial.values.every((v) => v !== "inf")
+    trial.values.every((v) => v !== "inf" && v !== "-inf")
   )
 }
 

optuna_dashboard/ts/components/GraphSlice.tsx

@@ -130,7 +130,9 @@ const filterFunc = (
     return false
   }
   return (
-    trial.values.length > objectiveId && trial.values[objectiveId] !== "inf"
+    trial.values.length > objectiveId &&
+    trial.values[objectiveId] !== "inf" &&
+    trial.values[objectiveId] !== "-inf"
   )
 }
 

optuna_dashboard/ts/components/StudyDetail.tsx

@@ -515,13 +515,18 @@ export const TrialTable: FC<{ studyDetail: StudyDetail | null }> = ({
 
         if (firstVal === secondVal) {
           return 0
-        } else if (firstVal && secondVal) {
-          return firstVal < secondVal ? 1 : -1
-        } else if (firstVal) {
+        }
+        if (firstVal === undefined) {
           return -1
-        } else {
+        } else if (secondVal === undefined) {
           return 1
         }
+        if (firstVal === "-inf" || secondVal === "inf") {
+          return 1
+        } else if (secondVal === "-inf" || firstVal === "inf") {
+          return -1
+        }
+        return firstVal < secondVal ? 1 : -1
       },
       toCellValue: (i) => {
         if (trials[i].values === undefined) {
@@ -542,13 +547,18 @@ export const TrialTable: FC<{ studyDetail: StudyDetail | null }> = ({
 
           if (firstVal === secondVal) {
             return 0
-          } else if (firstVal && secondVal) {
-            return firstVal < secondVal ? 1 : -1
-          } else if (firstVal) {
+          }
+          if (firstVal === undefined) {
             return -1
-          } else {
+          } else if (secondVal === undefined) {
             return 1
           }
+          if (firstVal === "-inf" || secondVal === "inf") {
+            return 1
+          } else if (secondVal === "-inf" || firstVal === "inf") {
+            return -1
+          }
+          return firstVal < secondVal ? 1 : -1
         },
         toCellValue: (i) => {
           if (trials[i].values === undefined) {
@@ -647,7 +657,29 @@ export const TrialTable: FC<{ studyDetail: StudyDetail | null }> = ({
   const collapseIntermediateValueColumns: DataGridColumn<TrialIntermediateValue>[] =
     [
       { field: "step", label: "Step", sortable: true },
-      { field: "value", label: "Value", sortable: true },
+      {
+        field: "value",
+        label: "Value",
+        sortable: true,
+        less: (firstEl, secondEl): number => {
+          const firstVal = firstEl.value
+          const secondVal = secondEl.value
+          if (firstVal === secondVal) {
+            return 0
+          }
+          if (firstVal === "nan") {
+            return -1
+          } else if (secondVal === "nan") {
+            return 1
+          }
+          if (firstVal === "-inf" || secondVal === "inf") {
+            return 1
+          } else if (secondVal === "-inf" || firstVal === "inf") {
+            return -1
+          }
+          return firstVal < secondVal ? 1 : -1
+        },
+      },
     ]
   const collapseAttrColumns: DataGridColumn<Attribute>[] = [
     { field: "key", label: "Key", sortable: true },

optuna_dashboard/ts/types/index.d.ts

@@ -7,6 +7,8 @@ declare const APP_BAR_TITLE: string
 declare const API_ENDPOINT: string
 declare const URL_PREFIX: string
 
+type TrialValueNumber = number | "inf" | "-inf"
+type TrialIntermediateValueNumber = number | "inf" | "-inf" | "nan"
 type TrialState = "Running" | "Complete" | "Pruned" | "Fail" | "Waiting"
 type StudyDirection = "maximize" | "minimize" | "not_set"
 type Distribution =
@@ -21,7 +23,7 @@ type Distribution =
 
 declare interface TrialIntermediateValue {
   step: number
-  value: number | "inf"
+  value: TrialIntermediateValueNumber
 }
 
 declare interface TrialParam {
@@ -55,7 +57,7 @@ declare interface Trial {
   study_id: number
   number: number
   state: TrialState
-  values?: (number | "inf")[]
+  values?: TrialValueNumber[]
   intermediate_values: TrialIntermediateValue[]
   datetime_start?: Date
   datetime_complete?: Date

visual_regression_test.py

@@ -1,6 +1,5 @@
 import argparse
 import asyncio
-import math
 import os
 import sys
 import threading
@@ -87,13 +86,15 @@ def objective_single_dynamic(trial: optuna.Trial) -> float:
 
     study.optimize(objective_single_dynamic, n_trials=50)
 
-    # Single objective study with 'inf' value
+    # Single objective study with 'inf', '-inf', or 'nan' value
     study = optuna.create_study(study_name="single-inf", storage=storage)
 
     def objective_single_inf(trial: optuna.Trial) -> float:
         x = trial.suggest_float("x", -10, 10)
-        if x > 0:
-            return math.inf
+        if trial.number % 3 == 0:
+            return float("inf")
+        elif trial.number % 3 == 1:
+            return float("-inf")
         else:
             return x**2
 
@@ -152,13 +153,19 @@ def objective_prune_without_report(trial: optuna.Trial) -> float:
 
     study.optimize(objective_prune_without_report, n_trials=100)
 
-    # Single objective pruned after reported 'inf' value
+    # Single objective pruned after reported 'inf', '-inf', or 'nan'
     study = optuna.create_study(study_name="single-inf-report", storage=storage)
 
     def objective_single_inf_report(trial: optuna.Trial) -> float:
         x = trial.suggest_float("x", -10, 10)
+        if trial.number % 3 == 0:
+            trial.report(float("inf"), 1)
+        elif trial.number % 3 == 1:
+            trial.report(float("-inf"), 1)
+        else:
+            trial.report(float("nan"), 1)
+
         if x > 0:
-            trial.report(math.inf, 1)
             raise optuna.TrialPruned()
         else:
             return x**2