Using tuple in CategoricalParameter

[HeinrichHaller]

Using tuple in CategoricalParameter

[xmatthias]

well it has - it's a simple limitation (i call it bug) in scikit-optimize (scikit-optimize/scikit-optimize#967).

There's never really been a solution to this (given that scikit-optimize is no longer maintained, that's no surprise) - so the solution essentially remains the same - use strings with a separator that you can easily split it again.

[HeinrichHaller]

Thank you very much, I came up with something like that as well. Quite handy addition to the usually fast expanding search space, that is not really inclined to converge to the global min.
sim = [(0.2, 0.1, 0.5), (0.3, 0.1, 0.7), (0.4, 0.1, 0.3)]
def format_number(n): if isinstance(n, float): # Format float to remove trailing zeros and keep up to 4 decimal places return f"{n:.4f}".rstrip('0').rstrip('.') elif isinstance(n, int): # Simply convert integer to string return str(n) sim_strings = [' '.join(format_number(x) for x in tuple) for tuple in sim]
I did not even know it is no longer maintained. Thanks for the quick heads up!

[xmatthias]

well we do keep it "installable" as part of a fork (otherwise some dependencies would not be updateable anymore) - but that's about it - there's no plans to add features, or fix additional bugs.

[HeinrichHaller]

Still get it not going..

@staticmethod
    def format_pair(pair):
        """Format the pair as a string with one decimal place."""
        return f"{pair[0]:.1f}_{pair[1]:.1f}"

# Use static method to format pairs
    simple_test_categories = [(0.2, 0.1), (0.3, 0.1), (0.4, 0.1)]
    formatted_categories = [Strategy.format_pair(pair) for pair in simple_test_categories]
    upper_lower_pairs = CategoricalParameter(formatted_categories, default="0.3_0.1", space='buy')

and then in populate_indicators:

def populate_indicators(self, dataframe, metadata):
        """Generate all indicators using parameters from CategoricalParameter."""
        # Iterate over each value in the range of categorical parameters
        for val in self.upper_lower_pairs.range:
            fast, slow = map(float, val.split('_'))  # Split the string and convert to float

afterwards the dynamic creation with f"" of the column names. It still leads to the error:

Epochs ━╸━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   24/1000   2% • 0:00:13 • 0:03:43
2024-05-02 18:26:54,735 - freqtrade - ERROR - Fatal exception!
Traceback (most recent call last):
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/freqtrade/main.py", line 42, in main
    return_code = args['func'](args)
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/freqtrade/commands/optimize_commands.py", line 107, in start_hyperopt
    hyperopt.start()
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/freqtrade/optimize/hyperopt.py", line 615, in start
    asked, is_random = self.get_asked_points(n_points=current_jobs)
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/freqtrade/optimize/hyperopt.py", line 503, in get_asked_points
    asked = unique_list(self.opt.ask(n_points=n_points * 5 if i > 0 else n_points))
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/optimizer/optimizer.py", line 395, in ask
    x = opt.ask()
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/optimizer/optimizer.py", line 367, in ask
    return self._ask()
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/optimizer/optimizer.py", line 446, in _ask
    min_delta_x = min([self.space.distance(next_x, xi)
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/optimizer/optimizer.py", line 446, in <listcomp>
    min_delta_x = min([self.space.distance(next_x, xi)
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/space/space.py", line 1138, in distance
    distance += dim.distance(a, b)
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/space/space.py", line 743, in distance
    if not (a in self and b in self):
  File "/home/moritz/VisualStudio/freqtrade5/freqtrade/.venv/lib/python3.10/site-packages/skopt/space/space.py", line 720, in __contains__
    return point in self.categories
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Could you maybe share how you do it? Thank you very much

[xmatthias]

well try it the simple way - hardcoding the strings instead of trying to overengineer things ?

i doubt the segments you shared will cause this though - at least not at first glance - so most likely you're loading a wrong strategy (duplicate strategies)...

[HeinrichHaller]

You are right, I was loading the wrong strategy. I did too many changes at once. I ended up with something like that:

def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        """Generate all indicators using parameters from CategoricalParameter."""
        print("Categorical parameter range:", self.upper_lower_pairs.range)
        print("Current value:", self.upper_lower_pairs.value)
        
        # Iterate over each value in the range of categorical parameters
        for val in self.upper_lower_pairs.range:
            print(f"Current val: {val}, type: {type(val)}")  # Debugging output
            fast, slow = map(float, val.split('_'))  # Split the string and convert to float

And the wrong datatype -list- still showed up. I restarted VS, restarted the .venv, ensured the pathes are correct.. I In the end I simply renamed the strategy and everything worked perfectly fine. I am such a noob.