__init__.py 7.79 KB
Newer Older
1
import os
2
import abc
lintangsutawika's avatar
lintangsutawika committed
3
import yaml
4
import collections
lintangsutawika's avatar
lintangsutawika committed
5

lintangsutawika's avatar
lintangsutawika committed
6
from functools import partial, lru_cache
7
from typing import List, Union, Dict
&'s avatar
& committed
8

9
from lm_eval import utils
10
from lm_eval import prompts
11
from lm_eval.api.task import TaskConfig, Task, ConfigurableTask
lintangsutawika's avatar
lintangsutawika committed
12

lintangsutawika's avatar
lintangsutawika committed
13
import logging
lintangsutawika's avatar
lintangsutawika committed
14

lintangsutawika's avatar
lintangsutawika committed
15
# import python tasks
lintangsutawika's avatar
lintangsutawika committed
16
from .squadv2.task import SQuAD2
lintangsutawika's avatar
lintangsutawika committed
17
18
19
20
21
22
23
24
from .scrolls.task import (
    QuALITY,
    NarrativeQA,
    ContractNLI,
    GovReport,
    SummScreenFD,
    QMSum,
)
lintangsutawika's avatar
lintangsutawika committed
25

26
eval_logger = utils.eval_logger
27

lintangsutawika's avatar
lintangsutawika committed
28

29
class TaskManager(abc.ABC):
30

31
    def __init__(
lintangsutawika's avatar
lintangsutawika committed
32
        self,
33
34
35
        verbosity="INFO",
        include_path=None
        ) -> None:
36

lintangsutawika's avatar
lintangsutawika committed
37
38
        self.verbosity = verbosity
        self.include_path = include_path
lintangsutawika's avatar
lintangsutawika committed
39
        self.logger = eval_logger.setLevel(getattr(logging, f"{verbosity}"))
lintangsutawika's avatar
lintangsutawika committed
40
41

        self.ALL_TASKS = self.initialize_tasks(
42
43
44
            include_path=include_path
            )

lintangsutawika's avatar
lintangsutawika committed
45
    def initialize_tasks(self, include_path=None):
lintangsutawika's avatar
lintangsutawika committed
46

lintangsutawika's avatar
lintangsutawika committed
47
48
49
50
51
52
53
54
        all_paths = [os.path.dirname(os.path.abspath(__file__)) + "/"]
        if include_path is not None:
            if isinstance(include_path, str):
                include_path = [include_path]
            all_paths.extend(include_path)

        ALL_TASKS = {}
        for task_dir in all_paths:
lintangsutawika's avatar
lintangsutawika committed
55
            tasks = self._get_task_and_group(task_dir)
lintangsutawika's avatar
lintangsutawika committed
56
57
58
59
            ALL_TASKS = {**tasks, **ALL_TASKS}

        return ALL_TASKS

60
    def all_tasks(self):
lintangsutawika's avatar
lintangsutawika committed
61
        return sorted(list(self.ALL_TASKS.keys()))
62

lintangsutawika's avatar
lintangsutawika committed
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
    def _name_is_registered(self, name):
        if name in self.ALL_TASKS:
            return True
        return False

    def _name_is_task(self, name):
        if self.ALL_TASKS[name]["type"] == "task":
            return True
        return False

    def _config_is_task(self, config):
        if list(config.keys()) == ["group", "task"]:
            return False
        return True

lintangsutawika's avatar
lintangsutawika committed
78
79
80
81
    def _get_yaml_path(self, name):
        assert name in self.ALL_TASKS
        return self.ALL_TASKS[name]["yaml_path"]

lintangsutawika's avatar
lintangsutawika committed
82
83
    def _get_config(self, name):
        assert name in self.ALL_TASKS
lintangsutawika's avatar
lintangsutawika committed
84
        yaml_path = self._get_yaml_path(name)
lintangsutawika's avatar
lintangsutawika committed
85
86
87
88
89
90
        return utils.load_yaml_config(yaml_path)

    def _get_tasklist(self, name):
        assert self._name_is_task(name) == False
        return self.ALL_TASKS[name]["task"]

lintangsutawika's avatar
lintangsutawika committed
91
    @lru_cache(None)
lintangsutawika's avatar
lintangsutawika committed
92
93
    def _load_individual_task_or_group(self, name_or_config: Union[str, dict] = None, parent_name: str = None) -> ConfigurableTask:

lintangsutawika's avatar
lintangsutawika committed
94
95
96
97
98
99
        def load_task(config, task, group=None):
            task_object = ConfigurableTask(config=config)
            if group is not None:
                task_object = (group, task_object)
            return {task: task_object}

lintangsutawika's avatar
lintangsutawika committed
100
101
102
        if isinstance(name_or_config, str):
            if self._name_is_task(name_or_config):
                task_config = self._get_config(name_or_config)
lintangsutawika's avatar
lintangsutawika committed
103
                return load_task(task_config, task=name_or_config, group=parent_name)
104
            else:
lintangsutawika's avatar
lintangsutawika committed
105
106
                group_name = name_or_config
                subtask_list = self._get_tasklist(name_or_config)
107
                if subtask_list == -1:
lintangsutawika's avatar
lintangsutawika committed
108
109
110
111
112
113
114
115
116
117
118
                    subtask_list = self._get_config(name_or_config)["task"]

        elif isinstance(name_or_config, dict):
            if self._config_is_task(name_or_config):
                task_name = name_or_config["task"]
                if self._name_is_registered(task_name):
                    base_task_config = self._get_config(task_name)
                    task_config={
                            **base_task_config,
                            **name_or_config,
                        }
119
                else:
lintangsutawika's avatar
lintangsutawika committed
120
                    task_config = name_or_config
lintangsutawika's avatar
lintangsutawika committed
121
                return load_task(task_config, task=name_or_config, group=parent_name)
122
            else:
lintangsutawika's avatar
lintangsutawika committed
123
124
125
                group_name = name_or_config["group"]
                subtask_list = name_or_config["task"]

lintangsutawika's avatar
lintangsutawika committed
126
127
128
129
130
        if self._get_yaml_path(group_name) == -1:
            all_subtasks = {group_name: (parent_name, None)}
        else:
            all_subtasks = {}

lintangsutawika's avatar
lintangsutawika committed
131
        fn = partial(self._load_individual_task_or_group, parent_name=group_name)
lintangsutawika's avatar
lintangsutawika committed
132
        all_subtasks = {**all_subtasks, **dict(collections.ChainMap(*map(fn, subtask_list)))}
lintangsutawika's avatar
lintangsutawika committed
133
134
        return all_subtasks

135
136
137
138
139
140

    def load_task_or_group(self, task_list: Union[str, list] = None) -> dict:

        if isinstance(task_list, str):
            task_list = [task_list]

lintangsutawika's avatar
lintangsutawika committed
141
142
143
144
145
146
        all_loaded_tasks = dict(
            collections.ChainMap(
                *map(
                    self._load_individual_task_or_group,
                    task_list
                )
147
            )
lintangsutawika's avatar
lintangsutawika committed
148
        )
149
        return all_loaded_tasks
150

lintangsutawika's avatar
lintangsutawika committed
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
    def _get_task_and_group(self, task_dir: str):
        tasks_and_groups = collections.defaultdict()
        for root, _, file_list in os.walk(task_dir):
            for f in file_list:
                if f.endswith(".yaml"):
                    yaml_path = os.path.join(root, f)
                    config = utils.simple_load_yaml_config(yaml_path)
                    if list(config.keys()) == ["group", "task"]:
                        # This is a group config
                        tasks_and_groups[config["group"]] = {
                            "type": "group",
                            "task": -1, # This signals that
                                        # we don't need to know
                                        # the task list for indexing
                                        # as it can be loaded
                                        # when called.
                            "yaml_path": yaml_path,
                        }
                    else:
                        # This is a task config
                        task = config["task"]
                        tasks_and_groups[task] = {
                            "type": "task",
                            "yaml_path": yaml_path,
                            }

                        if "group" in config:
                            groups = config["group"]
                            if isinstance(config["group"], str):
                                groups = [groups]

                            for group in groups:
                                if group not in tasks_and_groups:
                                    tasks_and_groups[group] = {
                                        "type": "group",
                                        "task": [task],
                                        "yaml_path": -1,
                                    }
                                else:
                                    tasks_and_groups[group]["task"].append(task)

        return tasks_and_groups
193

lintangsutawika's avatar
format  
lintangsutawika committed
194

lintangsutawika's avatar
lintangsutawika committed
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
# def check_prompt_config(
#     config: Dict[str, str], yaml_path: str = None
# ) -> List[Dict[str, str]]:
#     all_configs = []
#     if "use_prompt" in config:
#         prompt_list = prompts.load_prompt_list(
#             use_prompt=config["use_prompt"],
#             dataset_name=config["dataset_path"],
#             subset_name=config["dataset_name"] if "dataset_name" in config else None,
#             yaml_path=yaml_path,
#         )
#         for idx, prompt_variation in enumerate(prompt_list):
#             all_configs.append(
#                 {
#                     **config,
#                     **{"use_prompt": prompt_variation},
#                     **{
#                         "task": "_".join(
#                             [
#                                 config["task"]
#                                 if "task" in config
#                                 else get_task_name_from_config(config),
#                                 prompt_variation.split("/")[-1]
#                                 if ".yaml" in prompt_variation
#                                 else prompt_variation,
#                             ]
#                         )
#                     },
#                     **{"output_type": "generate_until"},
#                 }
#             )
#     else:
#         all_configs.append(config)
#     return all_configs