Skip to content

Random Search

Random search class.

BcoRandomSearch

Bases: BcoParameterSearch

BCO random search class. Subclass of BcoParameterSearch.

Source code in parameter_search/random_search.py 
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
class BcoRandomSearch(BcoParameterSearch):
    """BCO random search class. Subclass of `BcoParameterSearch`."""

    def __init__(self, search_space: SearchSpace, subset_size: int = 5):
        """Constructor.

        Parameters
        ----------
        search_space : SearchSpace
            The parameter search space.
        subset_size : int (default: 5)
            The number of parameter sets to search.
        """
        super().__init__(search_space)
        self.subset_size = subset_size

    def _setup_logger(
        self, path: str = "./logs", name: str = "random-search"
    ) -> Logger:
        """Sets up the logger.

        Parameters
        ----------
        path : str, optional
            File path for the logger.
        name : str, optional
            Name for the logger output.

        Returns
        -------
        Logger
            The grid search logger.
        """
        check_dir(path)
        if not name.endswith(".log"):
            name = f"{name}.log"
        return setup_root_logger(os.path.join(path, name))

    def _create_param_sets(self) -> list[UserSelections]:
        """Creates a random subset of the parameter space.

        Returns
        -------
        list[UserSelections]
            A random subset of the search space combinations.
        """
        param_sets: list[UserSelections] = []

        for (
            llm,
            embedding_model,
            filepath,
            loader,
            chunking_config,
            vector_store,
            similarity_top_k,
        ) in product(
            self._llms,
            self._embedding_models,
            self._files,
            self._loaders,
            self._chunking_configs,
            self._vector_stores,
            self._similarity_top_k,
        ):
            base_selections = {
                "llm": llm,
                "embedding_model": embedding_model,
                "filename": os.path.basename(str(filepath)),
                "filepath": filepath,
                "vector_store": vector_store,
                "loader": loader,
                "mode": "production",
                "similarity_top_k": similarity_top_k,
                "chunking_config": chunking_config,
            }

            if self._git_data is None:
                base_selections["git_data"] = None
            else:
                for git_data in self._git_data:
                    if git_data["filename"] == filepath or git_data[
                        "filename"
                    ] == os.path.basename(str(filepath)):
                        base_selections["git_data"] = create_git_data(
                            user=git_data["git_info"]["user"],
                            repo=git_data["git_info"]["repo"],
                            branch=git_data["git_info"]["branch"],
                            filters=git_data["git_info"]["filters"],
                        )

            if self._other_docs is None:
                base_selections["other_docs"] = None
            else:
                for paper, other_docs in self._other_docs.items():
                    if paper == os.path.basename(str(filepath)):
                        base_selections["other_docs"] = other_docs

            user_selections = create_user_selections(
                base_selections["llm"],
                base_selections["embedding_model"],
                base_selections["filename"],
                base_selections["filepath"],
                base_selections["vector_store"],
                base_selections["loader"],
                base_selections["mode"],
                base_selections["similarity_top_k"],
                base_selections["chunking_config"],
                base_selections["git_data"],
                base_selections["other_docs"],
            )
            param_sets.append(user_selections)

        if self.subset_size > len(param_sets):
            self.subset_size = len(param_sets)

        param_subset = random.sample(param_sets, self.subset_size)

        return param_subset

__init__(search_space, subset_size=5)

Constructor.

Parameters:

Name Type Description Default
search_space SearchSpace

The parameter search space.

 required
subset_size int (default: 5)

The number of parameter sets to search.

 5
Source code in parameter_search/random_search.py 
21
22
23
24
25
26
27
28
29
30
31
32
def __init__(self, search_space: SearchSpace, subset_size: int = 5):
    """Constructor.

    Parameters
    ----------
    search_space : SearchSpace
        The parameter search space.
    subset_size : int (default: 5)
        The number of parameter sets to search.
    """
    super().__init__(search_space)
    self.subset_size = subset_size

_setup_logger(path='./logs', name='random-search')

Sets up the logger.

Parameters:

Name Type Description Default
path str

File path for the logger.

 './logs'
name str

Name for the logger output.

 'random-search'

Returns:

Type Description
Logger

The grid search logger.
Source code in parameter_search/random_search.py 
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
def _setup_logger(
    self, path: str = "./logs", name: str = "random-search"
) -> Logger:
    """Sets up the logger.

    Parameters
    ----------
    path : str, optional
        File path for the logger.
    name : str, optional
        Name for the logger output.

    Returns
    -------
    Logger
        The grid search logger.
    """
    check_dir(path)
    if not name.endswith(".log"):
        name = f"{name}.log"
    return setup_root_logger(os.path.join(path, name))

_create_param_sets()

Creates a random subset of the parameter space.

Returns:

Type Description
list[UserSelections]

A random subset of the search space combinations.
Source code in parameter_search/random_search.py 
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
def _create_param_sets(self) -> list[UserSelections]:
    """Creates a random subset of the parameter space.

    Returns
    -------
    list[UserSelections]
        A random subset of the search space combinations.
    """
    param_sets: list[UserSelections] = []

    for (
        llm,
        embedding_model,
        filepath,
        loader,
        chunking_config,
        vector_store,
        similarity_top_k,
    ) in product(
        self._llms,
        self._embedding_models,
        self._files,
        self._loaders,
        self._chunking_configs,
        self._vector_stores,
        self._similarity_top_k,
    ):
        base_selections = {
            "llm": llm,
            "embedding_model": embedding_model,
            "filename": os.path.basename(str(filepath)),
            "filepath": filepath,
            "vector_store": vector_store,
            "loader": loader,
            "mode": "production",
            "similarity_top_k": similarity_top_k,
            "chunking_config": chunking_config,
        }

        if self._git_data is None:
            base_selections["git_data"] = None
        else:
            for git_data in self._git_data:
                if git_data["filename"] == filepath or git_data[
                    "filename"
                ] == os.path.basename(str(filepath)):
                    base_selections["git_data"] = create_git_data(
                        user=git_data["git_info"]["user"],
                        repo=git_data["git_info"]["repo"],
                        branch=git_data["git_info"]["branch"],
                        filters=git_data["git_info"]["filters"],
                    )

        if self._other_docs is None:
            base_selections["other_docs"] = None
        else:
            for paper, other_docs in self._other_docs.items():
                if paper == os.path.basename(str(filepath)):
                    base_selections["other_docs"] = other_docs

        user_selections = create_user_selections(
            base_selections["llm"],
            base_selections["embedding_model"],
            base_selections["filename"],
            base_selections["filepath"],
            base_selections["vector_store"],
            base_selections["loader"],
            base_selections["mode"],
            base_selections["similarity_top_k"],
            base_selections["chunking_config"],
            base_selections["git_data"],
            base_selections["other_docs"],
        )
        param_sets.append(user_selections)

    if self.subset_size > len(param_sets):
        self.subset_size = len(param_sets)

    param_subset = random.sample(param_sets, self.subset_size)

    return param_subset