__init__.py 26.9 KB
Newer Older
1
import collections
2
import inspect
3
4
import logging
import os
5
from functools import partial
6
from typing import Dict, List, Mapping, Optional, Union
&'s avatar
& committed
7

8
from lm_eval import utils
Lintang Sutawika's avatar
Lintang Sutawika committed
9
from lm_eval.api.group import ConfigurableGroup, GroupConfig
10
from lm_eval.api.task import ConfigurableTask, Task
Lintang Sutawika's avatar
Lintang Sutawika committed
11
12
13
14
from lm_eval.evaluator_utils import get_subtask_list


GROUP_ONLY_KEYS = list(GroupConfig().to_dict().keys())
lintangsutawika's avatar
lintangsutawika committed
15

16

Baber's avatar
test  
Baber committed
17
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
def convert_mcq_to_generative(cfg: dict):
    prompt = """Given the following question and candidate answers, choose the correct answer."""
    if cfg.get("output_type", "generate_until") == "generate_until":
        return cfg
    else:
        cfg["output_type"] = "generate_until"
        cfg["doc_to_text"] = (
            prompt
            + cfg.get("doc_to_text", "")
            + 'Your response should end with "The best answer is [the_answer_letter]" where the [the_answer_letter] is one of choice letters, A, B, C etc.'
        )
        cfg["generation_kwargs"] = ({"until": ["."], "max_gen_toks": 10},)
        cfg["filter_list"] = (
            [
                {
                    "name": "strict_match",
                    "filter": [
                        {"function": "remove_whitespace"},
                        {"function": "take_first"},
                    ],
                }
            ],
        )
    return cfg


# def convert_mcq_to_generative(cfg: dict):
#     Prompt = """Given the following question and candidate answers, choose the correct answer."""
#     if cfg.get("output_type", "generate_until") == "generate_until":
#         return cfg
#     else:
#         cfg["output_type"] = "generate_until"
#         doc_to_text: str = cfg.get("doc_to_text", "")
#         doc_to_choice = cfg.get("doc_to_choice")
#         assert doc_to_choice is not None, "doc_to_choice is required!"
#         if isinstance(doc_to_choice, str):
#             doc_to_choice = doc_to_choice.replace("{", "").replace("}", "")
#         if doc_to_text.lower().rfind("answer") != -1:
#             doc_to_text = doc_to_text[:doc_to_text.lower().rfind(r"answer")].strip()
#         elif doc_to_text.lower().rfind("a:") != -1:
#             doc_to_text = doc_to_text[:doc_to_text.lower().rfind(r"a:")].strip()
#
#         cfg['doc_to_text'] = (
#             f"{Prompt + '\n' + doc_to_text + '\n'}"
#             "{% set letters = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J'] %}"
#             f"{{% for choice in {doc_to_choice} %}}"
#             "{{letters[loop.index0]}}. {{choice}}" + "\n"
#             "{% endfor %}\n"
#             """Your response should end with \"The best answer is [the_answer_letter]\" where the [the_answer_letter] is one of the answer letters."""
#         )
#         del cfg["doc_to_choice"]
#         cfg["gen_prefix"] = "The answer is"
#
#     return cfg


73
74
75
class TaskManager:
    """TaskManager indexes all tasks from the default `lm_eval/tasks/`
    and an optional directory if provided.
76

77
78
    """

79
80
81
82
83
84
    def __init__(
        self,
        verbosity="INFO",
        include_path: Optional[Union[str, List]] = None,
        include_defaults: bool = True,
    ) -> None:
85
86
87
88
89
        self.verbosity = verbosity
        self.include_path = include_path
        self.logger = utils.eval_logger
        self.logger.setLevel(getattr(logging, f"{verbosity}"))

90
91
92
        self._task_index = self.initialize_tasks(
            include_path=include_path, include_defaults=include_defaults
        )
93
        self._all_tasks = sorted(list(self._task_index.keys()))
94

95
96
97
98
        self._all_groups = sorted(
            [x for x in self._all_tasks if self._task_index[x]["type"] == "group"]
        )
        self._all_subtasks = sorted(
99
100
101
102
103
            [
                x
                for x in self._all_tasks
                if self._task_index[x]["type"] in ["task", "python_task"]
            ]
104
105
106
107
108
        )
        self._all_tags = sorted(
            [x for x in self._all_tasks if self._task_index[x]["type"] == "tag"]
        )

109
        self.task_group_map = collections.defaultdict(list)
110

111
112
113
114
115
    def initialize_tasks(
        self,
        include_path: Optional[Union[str, List]] = None,
        include_defaults: bool = True,
    ):
116
        """Creates a dictionary of tasks index.
117

118
119
120
121
122
        :param include_path: Union[str, List] = None
            An additional path to be searched for tasks recursively.
            Can provide more than one such path as a list.
        :param include_defaults: bool = True
            If set to false, default tasks (those in lm_eval/tasks/) are not indexed.
123
124
125
        :return
            Dictionary of task names as key and task metadata
        """
126
127
128
129
        if include_defaults:
            all_paths = [os.path.dirname(os.path.abspath(__file__)) + "/"]
        else:
            all_paths = []
130
131
132
133
        if include_path is not None:
            if isinstance(include_path, str):
                include_path = [include_path]
            all_paths.extend(include_path)
134

135
136
137
138
        task_index = {}
        for task_dir in all_paths:
            tasks = self._get_task_and_group(task_dir)
            task_index = {**tasks, **task_index}
lintangsutawika's avatar
format  
lintangsutawika committed
139

140
141
142
143
144
145
        return task_index

    @property
    def all_tasks(self):
        return self._all_tasks

146
147
148
149
150
151
152
153
154
155
156
157
    @property
    def all_groups(self):
        return self._all_groups

    @property
    def all_subtasks(self):
        return self._all_subtasks

    @property
    def all_tags(self):
        return self._all_tags

158
159
160
161
    @property
    def task_index(self):
        return self._task_index

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
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
    def list_all_tasks(
        self, list_groups=True, list_tags=True, list_subtasks=True
    ) -> str:
        from pytablewriter import MarkdownTableWriter

        def sanitize_path(path):
            # don't print full path if we are within the lm_eval/tasks dir !
            # if we aren't though, provide the full path.
            if "lm_eval/tasks/" in path:
                return "lm_eval/tasks/" + path.split("lm_eval/tasks/")[-1]
            else:
                return path

        group_table = MarkdownTableWriter()
        group_table.headers = ["Group", "Config Location"]
        gt_values = []
        for g in self.all_groups:
            path = self.task_index[g]["yaml_path"]
            if path == -1:
                path = "---"
            else:
                path = sanitize_path(path)
            gt_values.append([g, path])
        group_table.value_matrix = gt_values

        tag_table = MarkdownTableWriter()
        tag_table.headers = ["Tag"]
        tag_table.value_matrix = [[t] for t in self.all_tags]

        subtask_table = MarkdownTableWriter()
        subtask_table.headers = ["Task", "Config Location", "Output Type"]
        st_values = []
        for t in self.all_subtasks:
            path = self.task_index[t]["yaml_path"]

            output_type = ""

            # read the yaml file to determine the output type
            if path != -1:
                config = utils.load_yaml_config(path, mode="simple")
                if "output_type" in config:
                    output_type = config["output_type"]
                elif (
                    "include" in config
                ):  # if no output type, check if there is an include with an output type
                    include_path = path.split("/")[:-1] + config["include"]
                    include_config = utils.load_yaml_config(include_path, mode="simple")
                    if "output_type" in include_config:
                        output_type = include_config["output_type"]

            if path == -1:
                path = "---"
            else:
                path = sanitize_path(path)
            st_values.append([t, path, output_type])
        subtask_table.value_matrix = st_values

        result = "\n"
        if list_groups:
            result += group_table.dumps() + "\n\n"
        if list_tags:
            result += tag_table.dumps() + "\n\n"
        if list_subtasks:
            result += subtask_table.dumps() + "\n\n"
        return result

228
    def match_tasks(self, task_list):
229
        return utils.pattern_match(task_list, self.all_tasks)
230

231
    def _name_is_registered(self, name) -> bool:
232
233
234
235
        if name in self.all_tasks:
            return True
        return False

236
    def _name_is_task(self, name) -> bool:
Lintang Sutawika's avatar
Lintang Sutawika committed
237
238
239
240
241
242
        if self._name_is_registered(name) and (self.task_index[name]["type"] == "task"):
            return True
        return False

    def _name_is_tag(self, name) -> bool:
        if self._name_is_registered(name) and (self.task_index[name]["type"] == "tag"):
243
244
245
            return True
        return False

246
    def _name_is_group(self, name) -> bool:
247
248
249
        if self._name_is_registered(name) and (
            self.task_index[name]["type"] == "group"
        ):
250
251
252
253
            return True
        return False

    def _name_is_python_task(self, name):
254
255
256
        if self._name_is_registered(name) and (
            self.task_index[name]["type"] == "python_task"
        ):
257
258
259
            return True
        return False

260
    def _config_is_task(self, config) -> bool:
261
262
263
264
        if ("task" in config) and isinstance(config["task"], str):
            return True
        return False

265
    def _config_is_group(self, config) -> bool:
266
267
268
269
        if ("task" in config) and isinstance(config["task"], list):
            return True
        return False

270
    def _config_is_python_task(self, config) -> bool:
271
272
273
274
275
        if "class" in config:
            return True
        return False

    def _get_yaml_path(self, name):
276
277
        if name not in self.task_index:
            raise ValueError
278
279
280
        return self.task_index[name]["yaml_path"]

    def _get_config(self, name):
281
282
        if name not in self.task_index:
            raise ValueError
283
284
285
286
287
288
289
        yaml_path = self._get_yaml_path(name)
        if yaml_path == -1:
            return {}
        else:
            return utils.load_yaml_config(yaml_path, mode="full")

    def _get_tasklist(self, name):
290
291
        if self._name_is_task(name):
            raise ValueError
292
293
294
295
296
297
298
299
300
301
302
        return self.task_index[name]["task"]

    def _process_alias(self, config, group=None):
        # If the group is not the same as the original
        # group which the group alias was intended for,
        # Set the group_alias to None instead.
        if ("group_alias" in config) and ("group" in config) and group is not None:
            if config["group"] != group:
                config["group_alias"] = None
        return config

303
304
305
306
307
308
309
310
    def _class_has_config_in_constructor(self, cls):
        constructor = getattr(cls, "__init__", None)
        return (
            "config" in inspect.signature(constructor).parameters
            if constructor
            else False
        )

311
    def _load_individual_task_or_group(
312
        self,
313
314
315
316
        name_or_config: Optional[Union[str, dict]] = None,
        parent_name: Optional[str] = None,
        update_config: Optional[dict] = None,
    ) -> Mapping:
Lintang Sutawika's avatar
Lintang Sutawika committed
317
        def _load_task(config, task):
318
            if "include" in config:
319
320
                config = {
                    **utils.load_yaml_config(
Lintang Sutawika's avatar
Lintang Sutawika committed
321
                        yaml_path=None,
322
323
                        yaml_config={"include": config.pop("include")},
                        mode="full",
324
325
326
                    ),
                    **config,
                }
327
            if self._config_is_python_task(config):
328
329
330
331
332
333
                if self._class_has_config_in_constructor(config["class"]):
                    task_object = config["class"](config=config)
                else:
                    task_object = config["class"]()
                if isinstance(task_object, ConfigurableTask):
                    # very scuffed: set task name here. TODO: fixme?
334
                    task_object.config.task = task
335
            else:
Baber's avatar
test  
Baber committed
336
                config = convert_mcq_to_generative(config)
337
                task_object = ConfigurableTask(config=config)
Lintang Sutawika's avatar
Lintang Sutawika committed
338

339
340
            return {task: task_object}

Lintang Sutawika's avatar
Lintang Sutawika committed
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
        def _get_group_and_subtask_from_config(config):
            group_name = ConfigurableGroup(config=config)
            subtask_list = []
            for task in group_name.config["task"]:
                if isinstance(task, str) and self._name_is_tag(task):
                    subtask_list.extend(self._get_tasklist(task))
                else:
                    subtask_list.append(task)
            return group_name, subtask_list

        def _process_group_config(config, update_config=None):
            if update_config is not None:
                config = {**config, **update_config}
            _update_config = {
                k: v for k, v in config.items() if k not in GROUP_ONLY_KEYS
            }
            if not bool(_update_config):
                _update_config = None

            group_config = {k: v for k, v in config.items() if k in GROUP_ONLY_KEYS}
            return group_config, _update_config

363
364
365
366
        if isinstance(name_or_config, str):
            if update_config is not None:
                # Process name_or_config as a dict instead
                name_or_config = {"task": name_or_config, **update_config}
Lintang Sutawika's avatar
Lintang Sutawika committed
367
368
369
            elif self._name_is_task(name_or_config) or self._name_is_python_task(
                name_or_config
            ):
370
                task_config = self._get_config(name_or_config)
Lintang Sutawika's avatar
Lintang Sutawika committed
371
                return _load_task(task_config, task=name_or_config)
372
            else:
373
374
375
                subtask_list = self._get_tasklist(name_or_config)
                if subtask_list == -1:
                    group_config = self._get_config(name_or_config)
Lintang Sutawika's avatar
Lintang Sutawika committed
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
                    group_config, update_config = _process_group_config(group_config)
                    group_name, subtask_list = _get_group_and_subtask_from_config(
                        group_config
                    )
                else:
                    if self._name_is_tag(name_or_config):
                        fn = partial(
                            self._load_individual_task_or_group,
                            update_config=name_or_config
                            if isinstance(name_or_config, dict)
                            else None,
                        )
                        return dict(
                            collections.ChainMap(*map(fn, reversed(subtask_list)))
                        )
                    else:
                        group_name = ConfigurableGroup(
                            config={"group": name_or_config, "task": subtask_list}
                        )
395

396
397
        if isinstance(name_or_config, dict):
            if self._config_is_task(name_or_config):
Lintang Sutawika's avatar
Lintang Sutawika committed
398
399
400
                name = name_or_config.pop("task")
                if update_config is not None:
                    name_or_config = {**name_or_config, **update_config}
401
402
                # If the name is registered as a group
                if self._name_is_group(name):
Lintang Sutawika's avatar
Lintang Sutawika committed
403
404
405
406
407
408
409
410
411
                    group_config = self._get_config(name)

                    group_config, update_config = _process_group_config(
                        group_config, name_or_config
                    )
                    group_name, subtask_list = _get_group_and_subtask_from_config(
                        group_config
                    )
                elif self._name_is_tag(name):
412
                    subtask_list = self._get_tasklist(name)
Lintang Sutawika's avatar
Lintang Sutawika committed
413
414
415
416
417
                    fn = partial(
                        self._load_individual_task_or_group,
                        update_config=name_or_config,
                    )
                    return dict(collections.ChainMap(*map(fn, reversed(subtask_list))))
418
419
420
421
422
423
                else:
                    if self._name_is_registered(name):
                        base_task_config = self._get_config(name)

                        # Check if this is a duplicate.
                        if parent_name is not None:
424
425
426
427
428
429
430
431
                            num_duplicate = len(
                                list(
                                    filter(
                                        lambda x: x.startswith(name),
                                        self.task_group_map[parent_name],
                                    )
                                )
                            )
432
433
434
435
                            if num_duplicate > 0:
                                name = f"{name}-{num_duplicate}"
                            self.task_group_map[parent_name].append(name)

436
437
438
439
                        task_config = {
                            **base_task_config,
                            **name_or_config,
                        }
440
441
                    else:
                        task_config = name_or_config
Lintang Sutawika's avatar
Lintang Sutawika committed
442
                    return _load_task(task_config, task=name)
443
            else:
Lintang Sutawika's avatar
Lintang Sutawika committed
444
445
446
447
                group_config, update_config = _process_group_config(name_or_config)
                group_name, subtask_list = _get_group_and_subtask_from_config(
                    group_config
                )
448

449
450
451
452
453
        fn = partial(
            self._load_individual_task_or_group,
            parent_name=group_name,
            update_config=update_config,
        )
Lintang Sutawika's avatar
Lintang Sutawika committed
454
455
        return {
            group_name: dict(collections.ChainMap(*map(fn, reversed(subtask_list))))
456
        }
457

458
    def load_task_or_group(self, task_list: Optional[Union[str, list]] = None) -> dict:
459
        """Loads a dictionary of task objects from a list
460

461
462
        :param task_list: Union[str, list] = None
            Single string or list of string of task names to be loaded
463

464
465
466
467
468
        :return
            Dictionary of task objects
        """
        if isinstance(task_list, str):
            task_list = [task_list]
469

470
        all_loaded_tasks = dict(
471
            collections.ChainMap(*map(self._load_individual_task_or_group, task_list))
472
473
474
475
476
477
478
        )
        return all_loaded_tasks

    def load_config(self, config: Dict):
        return self._load_individual_task_or_group(config)

    def _get_task_and_group(self, task_dir: str):
479
        """Creates a dictionary of tasks index with the following metadata,
Lintang Sutawika's avatar
Lintang Sutawika committed
480
        - `type`, that can be either `task`, `python_task`, `group` or `tags`.
481
482
            `task` refer to regular task configs, `python_task` are special
            yaml files that only consists of `task` and `class` parameters.
Lintang Sutawika's avatar
Lintang Sutawika committed
483
484
            `group` are group configs. `tags` are labels that can be assigned
            to tasks to assist in sorting and calling tasks of certain themes.
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
        - `yaml_path`, path to the yaml file. If the entry is a `group` that
            was configured through a task config, the yaml_path will be -1
            and all subtasks will be listed in `task` (see below)
        - `task`, reserved for entries with `type` as `group`. This will list
            all subtasks. When a group config is created (as opposed to task
            config having `group` parameter set), this will be set to -1 to
            avoid recursive indexing. The whole list of subtasks will be loaded
            at evaluation.

        :param task_dir: str
            A directory to check for tasks

        :return
            Dictionary of task names as key and task metadata
        """
500
501
502

        def _populate_tags_and_groups(config, task, tasks_and_groups, print_info):
            # TODO: remove group in next release
503
504
505
506
507
508
509
510
511
512
513
514
515
            if "tag" in config:
                attr_list = config["tag"]
                if isinstance(attr_list, str):
                    attr_list = [attr_list]

                for tag in attr_list:
                    if tag not in tasks_and_groups:
                        tasks_and_groups[tag] = {
                            "type": "tag",
                            "task": [task],
                            "yaml_path": -1,
                        }
                    elif tasks_and_groups[tag]["type"] != "tag":
516
                        self.logger.info(
517
518
                            f"The tag '{tag}' is already registered as a group, this tag will not be registered. "
                            "This may affect tasks you want to call."
519
                        )
520
521
522
                        break
                    else:
                        tasks_and_groups[tag]["task"].append(task)
523

Lintang Sutawika's avatar
Lintang Sutawika committed
524
525
        # TODO: remove group in next release
        print_info = True
526
527
528
529
        ignore_dirs = [
            "__pycache__",
            ".ipynb_checkpoints",
        ]
530
        tasks_and_groups = collections.defaultdict()
531
532
        for root, dirs, file_list in os.walk(task_dir):
            dirs[:] = [d for d in dirs if d not in ignore_dirs]
533
534
535
536
537
538
            for f in file_list:
                if f.endswith(".yaml"):
                    yaml_path = os.path.join(root, f)
                    config = utils.load_yaml_config(yaml_path, mode="simple")
                    if self._config_is_python_task(config):
                        # This is a python class config
539
540
                        task = config["task"]
                        tasks_and_groups[task] = {
541
542
543
                            "type": "python_task",
                            "yaml_path": yaml_path,
                        }
544
545
546
                        _populate_tags_and_groups(
                            config, task, tasks_and_groups, print_info
                        )
547
548
549
550
                    elif self._config_is_group(config):
                        # This is a group config
                        tasks_and_groups[config["group"]] = {
                            "type": "group",
551
552
553
554
555
                            "task": -1,  # This signals that
                            # we don't need to know
                            # the task list for indexing
                            # as it can be loaded
                            # when called.
556
557
                            "yaml_path": yaml_path,
                        }
lintangsutawika's avatar
lintangsutawika committed
558

559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
                        # # Registered the level 1 tasks from a group config
                        # for config in config["task"]:
                        #     if isinstance(config, dict) and self._config_is_task(config):
                        #         task = config["task"]
                        #         tasks_and_groups[task] = {
                        #             "type": "task",
                        #             "yaml_path": yaml_path,
                        #             }

                    elif self._config_is_task(config):
                        # This is a task config
                        task = config["task"]
                        tasks_and_groups[task] = {
                            "type": "task",
                            "yaml_path": yaml_path,
574
                        }
575
576
577
                        _populate_tags_and_groups(
                            config, task, tasks_and_groups, print_info
                        )
578
579
580
581
                    else:
                        self.logger.debug(f"File {f} in {root} could not be loaded")

        return tasks_and_groups
lintangsutawika's avatar
lintangsutawika committed
582

583

584
585
586
587
588
589
590
def get_task_name_from_config(task_config: Dict[str, str]) -> str:
    if "task" in task_config:
        return task_config["task"]
    if "dataset_name" in task_config:
        return "{dataset_path}_{dataset_name}".format(**task_config)
    else:
        return "{dataset_path}".format(**task_config)
lintangsutawika's avatar
lintangsutawika committed
591

592

lintangsutawika's avatar
lintangsutawika committed
593
def get_task_name_from_object(task_object):
594
595
    if hasattr(task_object, "config"):
        return task_object._config["task"]
lintangsutawika's avatar
lintangsutawika committed
596
597
598
599
600
601
602
603
604

    # TODO: scrap this
    # this gives a mechanism for non-registered tasks to have a custom name anyways when reporting
    return (
        task_object.EVAL_HARNESS_NAME
        if hasattr(task_object, "EVAL_HARNESS_NAME")
        else type(task_object).__name__
    )

605

Lintang Sutawika's avatar
Lintang Sutawika committed
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
def _check_duplicates(task_dict: dict) -> List[str]:
    """helper function solely used in validating get_task_dict output.
    Takes the output of lm_eval.evaluator_utils.get_subtask_list and
    returns a list of all leaf subtasks contained within, and errors if any such leaf subtasks are
    "oversubscribed" to several disjoint groups.
    """
    subtask_names = []
    for key, value in task_dict.items():
        subtask_names.extend(value)

    duplicate_tasks = {
        task_name for task_name in subtask_names if subtask_names.count(task_name) > 1
    }

    # locate the potentially problematic groups that seem to 'compete' for constituent subtasks
    competing_groups = [
        group
        for group in task_dict.keys()
        if len(set(task_dict[group]).intersection(duplicate_tasks)) > 0
    ]

    if len(duplicate_tasks) > 0:
        raise ValueError(
            f"Found 1 or more tasks while trying to call get_task_dict() that were members of more than 1 called group: {list(duplicate_tasks)}. Offending groups: {competing_groups}. Please call groups which overlap their constituent tasks in separate evaluation runs."
        )


633
def get_task_dict(
634
    task_name_list: Union[str, List[Union[str, Dict, Task]]],
635
    task_manager: Optional[TaskManager] = None,
636
):
637
    """Creates a dictionary of task objects from either a name of task, config, or prepared Task object.
lintangsutawika's avatar
lintangsutawika committed
638

639
640
641
642
643
644
645
    :param task_name_list: List[Union[str, Dict, Task]]
        Name of model or LM object, see lm_eval.models.get_model
    :param task_manager: TaskManager = None
        A TaskManager object that stores indexed tasks. If not set,
        task_manager will load one. This should be set by the user
        if there are additional paths that want to be included
        via `include_path`
646

647
648
649
    :return
        Dictionary of task objects
    """
Lintang Sutawika's avatar
Lintang Sutawika committed
650

651
    task_name_from_string_dict = {}
652
653
654
    task_name_from_config_dict = {}
    task_name_from_object_dict = {}

655
    if isinstance(task_name_list, str):
lintangsutawika's avatar
lintangsutawika committed
656
        task_name_list = [task_name_list]
657
658
659
660
661
662
663
664
665
    elif isinstance(task_name_list, list):
        if not all([isinstance(task, (str, dict, Task)) for task in task_name_list]):
            raise TypeError(
                "Expected all list items to be of types 'str', 'dict', or 'Task', but at least one entry did not match."
            )
    else:
        raise TypeError(
            f"Expected a 'str' or 'list' but received {type(task_name_list)}."
        )
lintangsutawika's avatar
lintangsutawika committed
666

667
    string_task_name_list = [task for task in task_name_list if isinstance(task, str)]
668
669
670
    others_task_name_list = [
        task for task in task_name_list if not isinstance(task, str)
    ]
671
672
673
    if len(string_task_name_list) > 0:
        if task_manager is None:
            task_manager = TaskManager()
lintangsutawika's avatar
lintangsutawika committed
674

675
676
677
        task_name_from_string_dict = task_manager.load_task_or_group(
            string_task_name_list
        )
678

679
680
    for task_element in others_task_name_list:
        if isinstance(task_element, dict):
681
682
            task_name_from_config_dict = {
                **task_name_from_config_dict,
683
                **task_manager.load_config(config=task_element),
684
685
686
687
688
            }

        elif isinstance(task_element, Task):
            task_name_from_object_dict = {
                **task_name_from_object_dict,
lintangsutawika's avatar
lintangsutawika committed
689
                get_task_name_from_object(task_element): task_element,
690
            }
lintangsutawika's avatar
lintangsutawika committed
691

692
    if not set(task_name_from_string_dict.keys()).isdisjoint(
lintangsutawika's avatar
lintangsutawika committed
693
        set(task_name_from_object_dict.keys())
694
695
    ):
        raise ValueError
696

Lintang Sutawika's avatar
Lintang Sutawika committed
697
    final_task_dict = {
698
        **task_name_from_string_dict,
lintangsutawika's avatar
lintangsutawika committed
699
        **task_name_from_config_dict,
700
        **task_name_from_object_dict,
lintangsutawika's avatar
lintangsutawika committed
701
    }
Lintang Sutawika's avatar
Lintang Sutawika committed
702
703
704
705
706
707
708
709

    # behavior can get odd if one tries to invoke several groups that "compete" for the same task.
    # (notably, because one could request several num_fewshot values at once in GroupConfig overrides for the subtask
    # and we'd be unsure which to use and report.)
    # we explicitly check and error in this case.
    _check_duplicates(get_subtask_list(final_task_dict))

    return final_task_dict