__init__.py 10.4 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
from lm_eval.api.registry import (
13
14
    register_task,
    register_group,
lintangsutawika's avatar
lintangsutawika committed
15
16
    TASK_REGISTRY,
    GROUP_REGISTRY,
lintangsutawika's avatar
lintangsutawika committed
17
)
lintangsutawika's avatar
lintangsutawika committed
18

lintangsutawika's avatar
lintangsutawika committed
19
import logging
lintangsutawika's avatar
lintangsutawika committed
20

lintangsutawika's avatar
lintangsutawika committed
21
# import python tasks
lintangsutawika's avatar
lintangsutawika committed
22
from .squadv2.task import SQuAD2
lintangsutawika's avatar
lintangsutawika committed
23
24
25
26
27
28
29
30
from .scrolls.task import (
    QuALITY,
    NarrativeQA,
    ContractNLI,
    GovReport,
    SummScreenFD,
    QMSum,
)
lintangsutawika's avatar
lintangsutawika committed
31

32
eval_logger = utils.eval_logger
33

lintangsutawika's avatar
lintangsutawika committed
34

35
class TaskManager(abc.ABC):
36

37
    def __init__(
lintangsutawika's avatar
lintangsutawika committed
38
        self,
39
40
41
        verbosity="INFO",
        include_path=None
        ) -> None:
42

lintangsutawika's avatar
lintangsutawika committed
43
44
        self.verbosity = verbosity
        self.include_path = include_path
lintangsutawika's avatar
lintangsutawika committed
45
        self.logger = eval_logger.setLevel(getattr(logging, f"{verbosity}"))
lintangsutawika's avatar
lintangsutawika committed
46
47

        self.ALL_TASKS = self.initialize_tasks(
48
49
50
            include_path=include_path
            )

lintangsutawika's avatar
lintangsutawika committed
51
    def initialize_tasks(self, include_path=None):
lintangsutawika's avatar
lintangsutawika committed
52

lintangsutawika's avatar
lintangsutawika committed
53
54
55
56
57
58
59
60
        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
61
            tasks = self._get_task_and_group(task_dir)
lintangsutawika's avatar
lintangsutawika committed
62
63
64
65
            ALL_TASKS = {**tasks, **ALL_TASKS}

        return ALL_TASKS

66
    def all_tasks(self):
lintangsutawika's avatar
lintangsutawika committed
67
        return sorted(list(self.ALL_TASKS.keys()))
68

lintangsutawika's avatar
lintangsutawika committed
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
    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
84
85
86
87
    def _get_yaml_path(self, name):
        assert name in self.ALL_TASKS
        return self.ALL_TASKS[name]["yaml_path"]

lintangsutawika's avatar
lintangsutawika committed
88
89
    def _get_config(self, name):
        assert name in self.ALL_TASKS
lintangsutawika's avatar
lintangsutawika committed
90
        yaml_path = self._get_yaml_path(name)
lintangsutawika's avatar
lintangsutawika committed
91
92
93
94
95
96
        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
97
    @lru_cache(None)
lintangsutawika's avatar
lintangsutawika committed
98
99
    def _load_individual_task_or_group(self, name_or_config: Union[str, dict] = None, parent_name: str = None) -> ConfigurableTask:

lintangsutawika's avatar
lintangsutawika committed
100
101
102
103
104
105
        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
106
107
108
        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
109
                return load_task(task_config, task=name_or_config, group=parent_name)
110
            else:
lintangsutawika's avatar
lintangsutawika committed
111
112
                group_name = name_or_config
                subtask_list = self._get_tasklist(name_or_config)
113
                if subtask_list == -1:
lintangsutawika's avatar
lintangsutawika committed
114
115
116
117
118
119
120
121
122
123
124
                    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,
                        }
125
                else:
lintangsutawika's avatar
lintangsutawika committed
126
                    task_config = name_or_config
lintangsutawika's avatar
lintangsutawika committed
127
                return load_task(task_config, task=name_or_config, group=parent_name)
128
            else:
lintangsutawika's avatar
lintangsutawika committed
129
130
131
                group_name = name_or_config["group"]
                subtask_list = name_or_config["task"]

lintangsutawika's avatar
lintangsutawika committed
132
133
134
135
136
        if self._get_yaml_path(group_name) == -1:
            all_subtasks = {group_name: (parent_name, None)}
        else:
            all_subtasks = {}

lintangsutawika's avatar
lintangsutawika committed
137
        fn = partial(self._load_individual_task_or_group, parent_name=group_name)
lintangsutawika's avatar
lintangsutawika committed
138
        all_subtasks = {**all_subtasks, **dict(collections.ChainMap(*map(fn, subtask_list)))}
lintangsutawika's avatar
lintangsutawika committed
139
140
        return all_subtasks

141
142
143
144
145
146

    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
147
148
149
150
151
152
        all_loaded_tasks = dict(
            collections.ChainMap(
                *map(
                    self._load_individual_task_or_group,
                    task_list
                )
153
            )
lintangsutawika's avatar
lintangsutawika committed
154
        )
155
        return all_loaded_tasks
156

lintangsutawika's avatar
lintangsutawika committed
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
193
194
195
196
197
198
    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
199

lintangsutawika's avatar
format  
lintangsutawika committed
200

lintangsutawika's avatar
lintangsutawika committed
201
202
203
def check_prompt_config(
    config: Dict[str, str], yaml_path: str = None
) -> List[Dict[str, str]]:
204
205
206
207
208
    all_configs = []
    if "use_prompt" in config:
        prompt_list = prompts.load_prompt_list(
            use_prompt=config["use_prompt"],
            dataset_name=config["dataset_path"],
lintangsutawika's avatar
lintangsutawika committed
209
            subset_name=config["dataset_name"] if "dataset_name" in config else None,
lintangsutawika's avatar
lintangsutawika committed
210
            yaml_path=yaml_path,
211
212
213
214
215
216
217
218
        )
        for idx, prompt_variation in enumerate(prompt_list):
            all_configs.append(
                {
                    **config,
                    **{"use_prompt": prompt_variation},
                    **{
                        "task": "_".join(
219
220
221
222
                            [
                                config["task"]
                                if "task" in config
                                else get_task_name_from_config(config),
lintangsutawika's avatar
lintangsutawika committed
223
224
225
                                prompt_variation.split("/")[-1]
                                if ".yaml" in prompt_variation
                                else prompt_variation,
226
227
228
                            ]
                        )
                    },
229
                    **{"output_type": "generate_until"},
230
231
232
233
234
235
                }
            )
    else:
        all_configs.append(config)
    return all_configs

lintangsutawika's avatar
lintangsutawika committed
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
# # TODO: pass num_fewshot and other cmdline overrides in a better way
# def get_task_dict(task_name_list: List[Union[str, Dict, Task]], **kwargs):
#     config = {**kwargs}

#     task_name_from_registry_dict = {}
#     task_name_from_config_dict = {}
#     task_name_from_object_dict = {}

#     if type(task_name_list) != list:
#         task_name_list = [task_name_list]

#     for task_element in task_name_list:
#         if isinstance(task_element, str):
#             if task_element in GROUP_REGISTRY:
#                 group_name = task_element
#                 for task_name in GROUP_REGISTRY[task_element]:
#                     if task_name not in task_name_from_registry_dict:
#                         task_obj = get_task_dict(task_name)
#                         if task_name in task_obj.keys():
#                             task_dict = {
#                                 task_name: (group_name, task_obj[task_name]),
#                             }
#                         else:
#                             task_dict = {
#                                 task_name: (group_name, None),
#                                 **task_obj,
#                             }

#                         task_name_from_registry_dict = {
#                             **task_name_from_registry_dict,
#                             **task_dict,
#                         }
#             else:
#                 task_name = task_element
#                 if task_name not in task_name_from_registry_dict:
#                     task_name_from_registry_dict = {
#                         **task_name_from_registry_dict,
#                         task_name: get_task(task_name=task_element, config=config),
#                     }

#         elif isinstance(task_element, dict):
#             task_element.update(config)
#             task_name_from_config_dict = {
#                 **task_name_from_config_dict,
#                 get_task_name_from_config(task_element): ConfigurableTask(
#                     config=task_element
#                 ),
#             }

#         elif isinstance(task_element, Task):
#             task_name_from_object_dict = {
#                 **task_name_from_object_dict,
#                 get_task_name_from_object(task_element): task_element,
#             }

#     assert set(task_name_from_registry_dict.keys()).isdisjoint(
#         set(task_name_from_object_dict.keys())
#     )
#     return {
#         **task_name_from_registry_dict,
#         **task_name_from_config_dict,
#         **task_name_from_object_dict,
#     }