__init__.py 8.85 KB
Newer Older
1
import os
lintangsutawika's avatar
lintangsutawika committed
2
import yaml
3
from typing import List, Union, Dict
&'s avatar
& committed
4

5
from lm_eval import utils
6
from lm_eval import prompts
lintangsutawika's avatar
lintangsutawika committed
7
from lm_eval.logger import eval_logger
8
from lm_eval.api.task import TaskConfig, Task, ConfigurableTask
lintangsutawika's avatar
lintangsutawika committed
9
from lm_eval.api.registry import (
10
11
    register_task,
    register_group,
lintangsutawika's avatar
lintangsutawika committed
12
13
    TASK_REGISTRY,
    GROUP_REGISTRY,
haileyschoelkopf's avatar
haileyschoelkopf committed
14
    ALL_TASKS,
lintangsutawika's avatar
lintangsutawika committed
15
)
lintangsutawika's avatar
lintangsutawika committed
16

17

18
def register_configurable_task(config: Dict[str, str]) -> int:
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
    SubClass = type(
        config["task"] + "ConfigurableTask",
        (ConfigurableTask,),
        {"CONFIG": TaskConfig(**config)},
    )

    if "task" in config:
        task_name = "{}".format(config["task"])
        register_task(task_name)(SubClass)

    if "group" in config:
        if type(config["group"]) == str:
            group_name = [config["group"]]
        else:
            group_name = config["group"]

        for group in group_name:
            register_group(group)(SubClass)

    return 0

lintangsutawika's avatar
format  
lintangsutawika committed
40

lintangsutawika's avatar
lintangsutawika committed
41
def register_configurable_group(config: Dict[str, str], yaml_path: str = None) -> int:
42
43
    group = config["group"]
    all_task_list = config["task"]
lintangsutawika's avatar
format  
lintangsutawika committed
44
45
    config_list = [task for task in all_task_list if type(task) != str]
    task_list = [task for task in all_task_list if type(task) == str]
46
47

    for task_config in config_list:
lintangsutawika's avatar
lintangsutawika committed
48
49
50
51
52
53
54
55
56
57
58
59
        # assert "task" in task_config:
        # task = task_config["task"]
        # if task in GROUP_REGISTRY:
        #     task_list = GROUP_REGISTRY[task]
        # elif task in TASK_REGISTRY:
        #     task_list = [TASK_REGISTRY[task]]

        # for _task in task_list:
        #     task_config = {
        #         **_task["CONFIG"],
        #         **task_config
        #     }
lintangsutawika's avatar
lintangsutawika committed
60
        task_config = utils.load_yaml_config(yaml_path, task_config)
61
62
63
64
        var_configs = check_prompt_config(
            {
                **task_config,
                **{"group": group},
lintangsutawika's avatar
lintangsutawika committed
65
66
            },
            yaml_path=os.path.dirname(yaml_path),
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
        )
        for config in var_configs:
            register_configurable_task(config)

    task_names = utils.pattern_match(task_list, ALL_TASKS)
    for task in task_names:
        if (task in TASK_REGISTRY) or (task in GROUP_REGISTRY):
            if group in GROUP_REGISTRY:
                GROUP_REGISTRY[group].append(task)
            else:
                GROUP_REGISTRY[group] = [task]
                ALL_TASKS.add(group)

    return 0

82

lintangsutawika's avatar
lintangsutawika committed
83
84
85
def check_prompt_config(
    config: Dict[str, str], yaml_path: str = None
) -> List[Dict[str, str]]:
86
87
88
89
90
    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
91
            subset_name=config["dataset_name"] if "dataset_name" in config else None,
lintangsutawika's avatar
lintangsutawika committed
92
            yaml_path=yaml_path,
93
94
95
96
97
98
99
100
        )
        for idx, prompt_variation in enumerate(prompt_list):
            all_configs.append(
                {
                    **config,
                    **{"use_prompt": prompt_variation},
                    **{
                        "task": "_".join(
101
102
103
104
                            [
                                config["task"]
                                if "task" in config
                                else get_task_name_from_config(config),
lintangsutawika's avatar
lintangsutawika committed
105
106
107
                                prompt_variation.split("/")[-1]
                                if ".yaml" in prompt_variation
                                else prompt_variation,
108
109
110
111
112
113
114
115
116
117
118
                            ]
                        )
                    },
                    **{"output_type": "greedy_until"},
                }
            )
    else:
        all_configs.append(config)
    return all_configs


119
def get_task_name_from_config(task_config: Dict[str, str]) -> str:
Lintang Sutawika's avatar
Lintang Sutawika committed
120
    if "dataset_name" in task_config:
Lintang Sutawika's avatar
Lintang Sutawika committed
121
122
123
        return "{dataset_path}_{dataset_name}".format(**task_config)
    else:
        return "{dataset_path}".format(**task_config)
124
125


lintangsutawika's avatar
lintangsutawika committed
126
def include_task_folder(task_dir: str, register_task: bool = True) -> None:
127
128
129
130
    """
    Calling this function
    """
    for root, subdirs, file_list in os.walk(task_dir):
131
132
133
134
135
136
137
        # if (subdirs == [] or subdirs == ["__pycache__"]) and (len(file_list) > 0):
        for f in file_list:
            if f.endswith(".yaml"):
                yaml_path = os.path.join(root, f)
                try:
                    config = utils.load_yaml_config(yaml_path)

lintangsutawika's avatar
lintangsutawika committed
138
139
                    if "task" not in config:
                        continue
lintangsutawika's avatar
lintangsutawika committed
140

lintangsutawika's avatar
lintangsutawika committed
141
142
143
144
145
146
147
148
149
150
                    all_configs = check_prompt_config(
                        config, yaml_path=os.path.dirname(yaml_path)
                    )
                    for config in all_configs:
                        if register_task:
                            if type(config["task"]) == str:
                                register_configurable_task(config)
                        else:
                            if type(config["task"]) == list:
                                register_configurable_group(config, yaml_path)
151
152

                except Exception as error:
lintangsutawika's avatar
lintangsutawika committed
153
                    import traceback
lintangsutawika's avatar
lintangsutawika committed
154
155
156

                    print("###")
                    print(yaml_path)
157
158
159
160
                    eval_logger.warning(
                        "Failed to load config in\n"
                        f"                                 {yaml_path}\n"
                        "                                 Config will not be added to registry\n"
lintangsutawika's avatar
lintangsutawika committed
161
162
                        f"                                 Error: {error}\n"
                        f"                                 Traceback: {traceback.format_exc()}"
163
                    )
lintangsutawika's avatar
lintangsutawika committed
164
    return 0
165
166
167
168
169
170
171


def include_path(task_dir):
    include_task_folder(task_dir)
    # Register Benchmarks after all tasks have been added
    include_task_folder(task_dir, register_task=False)
    return 0
172

lintangsutawika's avatar
lintangsutawika committed
173

174
task_dir = os.path.dirname(os.path.abspath(__file__)) + "/"
175
include_path(task_dir)
lintangsutawika's avatar
lintangsutawika committed
176

lintangsutawika's avatar
lintangsutawika committed
177

178
179
def get_task(task_name, config):
    try:
180
        return TASK_REGISTRY[task_name](config=config)
181
    except KeyError:
lintangsutawika's avatar
lintangsutawika committed
182
        eval_logger.info("Available tasks:")
183
        eval_logger.info(list(TASK_REGISTRY) + list(GROUP_REGISTRY))
lintangsutawika's avatar
lintangsutawika committed
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
        raise KeyError(f"Missing task {task_name}")


def get_task_name_from_object(task_object):
    for name, class_ in TASK_REGISTRY.items():
        if class_ is task_object:
            return name

    # 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__
    )


# TODO: pass num_fewshot and other cmdline overrides in a better way
202
def get_task_dict(task_name_list: List[Union[str, Dict, Task]], **kwargs):
lintangsutawika's avatar
lintangsutawika committed
203
204

    config = {**kwargs}
205
206
207
208
209

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

lintangsutawika's avatar
lintangsutawika committed
210
211
212
    if type(task_name_list) != list:
        task_name_list = [task_name_list]

213
214
215
216
    for task_element in task_name_list:
        if isinstance(task_element, str):

            if task_element in GROUP_REGISTRY:
217
                group_name = task_element
218
219
                for task_name in GROUP_REGISTRY[task_element]:
                    if task_name not in task_name_from_registry_dict:
lintangsutawika's avatar
lintangsutawika committed
220
221
222
223
224
225
226
227
228
229
230
                        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,
                            }

231
232
                        task_name_from_registry_dict = {
                            **task_name_from_registry_dict,
lintangsutawika's avatar
lintangsutawika committed
233
                            **task_dict,
lintangsutawika's avatar
lintangsutawika committed
234
                        }
235
            else:
236
                task_name = task_element
237
238
239
                if task_name not in task_name_from_registry_dict:
                    task_name_from_registry_dict = {
                        **task_name_from_registry_dict,
lintangsutawika's avatar
lintangsutawika committed
240
241
                        task_name: get_task(task_name=task_element, config=config),
                    }
242
243

        elif isinstance(task_element, dict):
244
            task_element.update(config)
245
246
247
            task_name_from_config_dict = {
                **task_name_from_config_dict,
                get_task_name_from_config(task_element): ConfigurableTask(
248
                    config=task_element
lintangsutawika's avatar
lintangsutawika committed
249
                ),
250
251
252
253
254
255
            }

        elif isinstance(task_element, Task):

            task_name_from_object_dict = {
                **task_name_from_object_dict,
lintangsutawika's avatar
lintangsutawika committed
256
                get_task_name_from_object(task_element): task_element,
257
            }
lintangsutawika's avatar
lintangsutawika committed
258
259
260
261

    assert set(task_name_from_registry_dict.keys()).isdisjoint(
        set(task_name_from_object_dict.keys())
    )
lintangsutawika's avatar
lintangsutawika committed
262
263
264
    return {
        **task_name_from_registry_dict,
        **task_name_from_config_dict,
265
        **task_name_from_object_dict,
lintangsutawika's avatar
lintangsutawika committed
266
    }