Error with `gp_minimize` in Hyper-parameter Optimization Chapter

[cleysonl]

Hi, I am running the code for gp_minimize.py and in the following part of the code I get an error:

result = gp_minimize(
        optimization_function,
        dimensions=param_space,
        n_calls=15,
        n_random_starts=10,
        verbose=10)

The error is the following: Exception has occurred: TypeError '<' not supported between instances of 'Version' and 'tuple'

Going through the gp_minimize info there is an example:

res = gp_minimize(f,                  # the function to minimize
                  [(-2.0, 2.0)],      # the bounds on each dimension of x
                  acq_func="EI",      # the acquisition function
                  n_calls=15,         # the number of evaluations of f
                  n_random_starts=5,  # the number of random initialization points
                  noise=0.1**2,       # the noise level (optional)
                  random_state=1234)   # the random seed

Where [(-2.0, 2.0)] is what we have as param_space in the book but if I print the latter we get: [Integer(low=3, high=15, prior='uniform', transform='identity'), Integer(low=100, high=1500, prior='uniform', transform='identity'), Categorical(categories=('gini', 'entropy'), prior=None), Real(low=0.01, high=1, prior='uniform', transform='identity')] Wondering if the error is the way we are passing the values but didn't found anything.

[abhishekkrthakur]

Are you using the versions mentioned in this git repo?

[cleysonl]

Good point, I am using windows so maybe that's the problem. Let me go through it again using Linux and the exact same versions in the git repo and I'll come back to you.

[abhishekkrthakur]

If im not wrong, windows shouldn't be a problem. could you try creating environment from files mentioned in this repo? README file, in the end.?

[cleysonl]

Used the requirements.txt in the repo for my conda environment and still got the error.

(ml) C:\Users\leyso\Documents\Courses\Approaching almost all ML problems\Chapter 8\src>python rf_gp_minimize.py
[Integer(low=3, high=15, prior='uniform', transform='identity'), Integer(low=100, high=1500, prior='uniform', transform='identity'), Categorical(categories=('gini', 'entropy'), prior=None), Real(low=0.01, high=1, prior='uniform', transform='identity')]
Iteration No: 1 started. Evaluating function at random point.
Traceback (most recent call last):
  File "rf_gp_minimize.py", line 109, in <module>
    result = gp_minimize(
  File "C:\Users\leyso\anaconda3\envs\ml\lib\site-packages\skopt\optimizer\gp.py", line 264, in gp_minimize
    return base_minimize(
  File "C:\Users\leyso\anaconda3\envs\ml\lib\site-packages\skopt\optimizer\base.py", line 271, in base_minimize
    next_x = optimizer.ask()
  File "C:\Users\leyso\anaconda3\envs\ml\lib\site-packages\skopt\optimizer\optimizer.py", line 332, in ask
    return self._ask()
  File "C:\Users\leyso\anaconda3\envs\ml\lib\site-packages\skopt\optimizer\optimizer.py", line 398, in _ask
    return self.space.rvs(random_state=self.rng)[0]
  File "C:\Users\leyso\anaconda3\envs\ml\lib\site-packages\skopt\space\space.py", line 764, in rvs
    if sp_version < (0, 16):
TypeError: '<' not supported between instances of 'Version' and 'tuple'

[cleysonl]

Had to install scikit-optimize because it wasn't in the file requirements.txt. Which version are you using? I have 0.7.4

[cleysonl]

Just ran the exactly same code with a new environment created inside WSL2 using the requirements.txt file in the repo an runs perfectly. Maybe it is a Windows thing.