fix optuna userwarning the range is not divisible by step

freqtrade/optimize/hyperopt/hyperopt.py

@@ -331,6 +331,7 @@ class Hyperopt:
                 self.config,
                 self.hyperopter.get_strategy_name(),
                 self.current_best_epoch,
+                self.hyperopter.o_dimensions,
             )
 
             HyperoptTools.show_epoch_details(

freqtrade/optimize/hyperopt_tools.py

@@ -7,6 +7,7 @@ from typing import Any
 
 import numpy as np
 import rapidjson
+from optuna.distributions import BaseDistribution
 from pandas import isna, json_normalize
 
 from freqtrade.constants import FTHYPT_FILEVERSION, Config
@@ -14,6 +15,7 @@ from freqtrade.enums import HyperoptState
 from freqtrade.exceptions import OperationalException
 from freqtrade.misc import deep_merge_dicts, round_dict, safe_value_fallback2
 from freqtrade.optimize.hyperopt_epoch_filters import hyperopt_filter_epochs
+from freqtrade.optimize.space import SKDecimal, _adjust_discrete_uniform
 
 
 logger = logging.getLogger(__name__)
@@ -61,11 +63,26 @@ class HyperoptTools:
         return None
 
     @staticmethod
-    def export_params(params, strategy_name: str, filename: Path):
+    def export_params(
+        params,
+        strategy_name: str,
+        filename: Path,
+        o_dimensions: dict[str, BaseDistribution] | None = None,
+    ):
         """
         Generate files
         """
         final_params = deepcopy(params["params_not_optimized"])
+        if o_dimensions:
+            for key, val in params["params_details"].items():
+                if isinstance(val, dict):
+                    for key1, val1 in val.items():
+                        if isinstance(o_dimensions.get(key1), SKDecimal):
+                            step = getattr(o_dimensions.get(key1), "step", None)
+                            if step:
+                                params["params_details"][key][key1] = _adjust_discrete_uniform(
+                                    val1, step
+                                )
         final_params = deep_merge_dicts(params["params_details"], final_params)
         final_params = {
             "strategy_name": strategy_name,
@@ -73,6 +90,7 @@ class HyperoptTools:
             "ft_stratparam_v": 1,
             "export_time": datetime.now(timezone.utc),
         }
+
         logger.info(f"Dumping parameters to {filename}")
         with filename.open("w") as f:
             rapidjson.dump(
@@ -93,12 +111,19 @@ class HyperoptTools:
         return params
 
     @staticmethod
-    def try_export_params(config: Config, strategy_name: str, params: dict):
+    def try_export_params(
+        config: Config,
+        strategy_name: str,
+        params: dict,
+        o_dimensions: dict[str, BaseDistribution] | None = None,
+    ):
         if params.get(FTHYPT_FILEVERSION, 1) >= 2 and not config.get("disableparamexport", False):
             # Export parameters ...
             fn = HyperoptTools.get_strategy_filename(config, strategy_name)
             if fn:
-                HyperoptTools.export_params(params, strategy_name, fn.with_suffix(".json"))
+                HyperoptTools.export_params(
+                    params, strategy_name, fn.with_suffix(".json"), o_dimensions
+                )
             else:
                 logger.warning("Strategy not found, not exporting parameter file.")
 

freqtrade/optimize/space/__init__.py

@@ -1,4 +1,4 @@
-from .decimalspace import SKDecimal
+from .decimalspace import SKDecimal, _adjust_discrete_uniform
 from .optunaspaces import (
     DimensionProtocol,
     ft_CategoricalDistribution,
@@ -13,10 +13,4 @@ Categorical = ft_CategoricalDistribution
 Integer = ft_IntDistribution
 Real = ft_FloatDistribution
 
-__all__ = [
+__all__ = ["Categorical", "Dimension", "Integer", "Real", "SKDecimal", "_adjust_discrete_uniform"]
-    "Categorical",
-    "Dimension",
-    "Integer",
-    "Real",
-    "SKDecimal",
-]

freqtrade/optimize/space/decimalspace.py

@@ -1,3 +1,5 @@
+import decimal
+
 from optuna.distributions import FloatDistribution
 
 
@@ -20,7 +22,32 @@ class SKDecimal(FloatDistribution):
         self.name = name
 
         super().__init__(
-            low=low,
+            low=_adjust_discrete_uniform(low, self.step),
-            high=high,
+            high=_adjust_discrete_uniform_high(low, high, self.step),
             step=self.step,
         )
+
+
+def _adjust_discrete_uniform_high(low: float, high: float, step: float | None) -> float:
+    if step:
+        d_high = decimal.Decimal(str(high))
+        d_low = decimal.Decimal(str(low))
+        d_step = decimal.Decimal(str(step))
+
+        d_r = d_high - d_low
+
+        if d_r % d_step != decimal.Decimal("0"):
+            high = float((d_r // d_step) * d_step + d_low)
+
+    return high
+
+
+def _adjust_discrete_uniform(val: float, step: float | None) -> float:
+    if step:
+        d_val = decimal.Decimal(str(val))
+        d_step = decimal.Decimal(str(step))
+
+        if d_val % d_step != decimal.Decimal("0"):
+            val = float((d_val // d_step) * d_step)
+
+    return val