__init__.py 13.4 KB
Newer Older
&'s avatar
& committed
1
from pprint import pprint
2
from typing import List, Union
&'s avatar
& committed
3

&'s avatar
& committed
4
import sacrebleu
5
import lm_eval.base
&'s avatar
& committed
6

Jason Phang's avatar
Jason Phang committed
7
8
from . import superglue
from . import glue
Leo Gao's avatar
Leo Gao committed
9
from . import arc
thefazzer's avatar
thefazzer committed
10
from . import coqa
Leo Gao's avatar
Leo Gao committed
11
from . import race
Leo Gao's avatar
Leo Gao committed
12
from . import webqs
Leo Gao's avatar
Leo Gao committed
13
from . import anli
14
from . import wsc273
Charles Foster's avatar
Charles Foster committed
15
from . import winogrande
Charles Foster's avatar
Charles Foster committed
16
from . import quac
Charles Foster's avatar
Charles Foster committed
17
from . import hellaswag
jon-tow's avatar
jon-tow committed
18
from . import swag
Charles Foster's avatar
Charles Foster committed
19
from . import openbookqa
Charles Foster's avatar
Charles Foster committed
20
from . import squad
21
from . import naturalqs
22
from . import sat
23
from . import arithmetic
Leo Gao's avatar
Leo Gao committed
24
from . import lambada
Leo Gao's avatar
Leo Gao committed
25
from . import piqa
Jonathan Tow's avatar
Jonathan Tow committed
26
from . import prost
Jonathan Tow's avatar
Jonathan Tow committed
27
from . import mc_taco
28
from . import triviaqa
jeffhsu3's avatar
jeffhsu3 committed
29
from . import pubmedqa
30
from . import sciq
31
from . import qasper
32
from . import qa4mre
&'s avatar
& committed
33
from . import translation
34
35
from . import headqa
from . import mathqa
Leo Gao's avatar
Leo Gao committed
36
from . import hendrycks_ethics
Jon Tow's avatar
Jon Tow committed
37
from . import drop
38
from . import unscramble
39
from . import logiqa
Andy Zou's avatar
Andy Zou committed
40
from . import hendrycks_test
Leo Gao's avatar
Leo Gao committed
41
from . import hendrycks_math
42
from . import cbt
43
from . import lambada_cloze
Jason Phang's avatar
Jason Phang committed
44
from . import pile
Leo Gao's avatar
Leo Gao committed
45
from . import wikitext
sdtblck's avatar
sdtblck committed
46
from . import lambada_multilingual
Jonathan Tow's avatar
Jonathan Tow committed
47
from . import mutual
48
from . import truthfulqa
Igor Ostrovsky's avatar
Igor Ostrovsky committed
49
from . import blimp
rokosbasilisk's avatar
rokosbasilisk committed
50
from . import asdiv
Jonathan Tow's avatar
Jonathan Tow committed
51
from . import gsm8k
Jonathan Tow's avatar
Jonathan Tow committed
52
from . import storycloze
53

&'s avatar
& committed
54
55
56
57
58
59
########################################
# Translation tasks
########################################

# 6 total
gpt3_translation_benchmarks = {
Fabrizio Milo's avatar
Fabrizio Milo committed
60
61
    "wmt14": ["en-fr", "fr-en"],  # French
    "wmt16": ["en-ro", "ro-en", "de-en", "en-de"],  # German, Romanian
&'s avatar
& committed
62
63
64
65
66
67
68
}


# 28 total
selected_translation_benchmarks = {
    **gpt3_translation_benchmarks,
    "wmt20": sacrebleu.get_langpairs_for_testset("wmt20"),
Fabrizio Milo's avatar
Fabrizio Milo committed
69
    "iwslt17": ["en-ar", "ar-en"],  # Arabic
&'s avatar
& committed
70
71
72
73
74
75
76
77
78
79
80
81
82
83
}

# 319 total
all_translation_benchmarks = {
    ts: sacrebleu.get_langpairs_for_testset(ts)
    for ts in sacrebleu.get_available_testsets()
}


########################################
# All tasks
########################################


Jason Phang's avatar
Jason Phang committed
84
TASK_REGISTRY = {
Jason Phang's avatar
multirc  
Jason Phang committed
85
    # GLUE
Jason Phang's avatar
Jason Phang committed
86
87
    "cola": glue.CoLA,
    "mnli": glue.MNLI,
Jason Phang's avatar
Jason Phang committed
88
    "mnli_mismatched": glue.MNLIMismatched,
Jason Phang's avatar
Jason Phang committed
89
90
91
92
    "mrpc": glue.MRPC,
    "rte": glue.RTE,
    "qnli": glue.QNLI,
    "qqp": glue.QQP,
Fabrizio Milo's avatar
Fabrizio Milo committed
93
    # "stsb": glue.STSB, # not implemented yet
Jason Phang's avatar
Jason Phang committed
94
95
    "sst": glue.SST,
    "wnli": glue.WNLI,
Jason Phang's avatar
multirc  
Jason Phang committed
96
    # SuperGLUE
Jason Phang's avatar
Jason Phang committed
97
    "boolq": superglue.BoolQ,
thefazzer's avatar
thefazzer committed
98
    "cb": superglue.CommitmentBank,
Jason Phang's avatar
Jason Phang committed
99
    "copa": superglue.Copa,
Jason Phang's avatar
multirc  
Jason Phang committed
100
    "multirc": superglue.MultiRC,
Jason Phang's avatar
Jason Phang committed
101
    "record": superglue.ReCoRD,
Jason Phang's avatar
Jason Phang committed
102
    "wic": superglue.WordsInContext,
Jason Phang's avatar
wsc  
Jason Phang committed
103
    "wsc": superglue.SGWinogradSchemaChallenge,
Jason Phang's avatar
multirc  
Jason Phang committed
104
    # Order by benchmark/genre?
thefazzer's avatar
thefazzer committed
105
    "coqa": coqa.CoQA,
Jon Tow's avatar
Jon Tow committed
106
    "drop": drop.DROP,
Leo Gao's avatar
Leo Gao committed
107
    "lambada": lambada.LAMBADA,
108
    "lambada_cloze": lambada_cloze.LAMBADA_cloze,
109
    # multilingual lambada
sdtblck's avatar
sdtblck committed
110
    **lambada_multilingual.construct_tasks(),
Leo Gao's avatar
Leo Gao committed
111
    "wikitext": wikitext.WikiText,
Leo Gao's avatar
Leo Gao committed
112
113
    # "cbt-cn": cbt.CBTCN, # disabled pending context length fix
    # "cbt-ne": cbt.CBTNE, # disabled pending context length fix
Leo Gao's avatar
Leo Gao committed
114
    "piqa": piqa.PiQA,
Jonathan Tow's avatar
Jonathan Tow committed
115
    "prost": prost.PROST,
Jonathan Tow's avatar
Jonathan Tow committed
116
    "mc_taco": mc_taco.MCTACO,
117
    # Science related
Fabrizio Milo's avatar
Fabrizio Milo committed
118
119
    "pubmedqa": pubmedqa.Pubmed_QA,
    "sciq": sciq.SciQ,
120
    "qasper": qasper.QASPER,
Fabrizio Milo's avatar
Fabrizio Milo committed
121
122
123
    "qa4mre_2011": qa4mre.QA4MRE_2011,
    "qa4mre_2012": qa4mre.QA4MRE_2012,
    "qa4mre_2013": qa4mre.QA4MRE_2013,
Leo Gao's avatar
Leo Gao committed
124
    "triviaqa": triviaqa.TriviaQA,
Jonathan Tow's avatar
Jonathan Tow committed
125
126
    "arc_easy": arc.ARCEasy,
    "arc_challenge": arc.ARCChallenge,
127
    # "quac": quac.QuAC, # not implemented yet
128
    "logiqa": logiqa.LogiQA,
Leo Gao's avatar
Leo Gao committed
129
    "hellaswag": hellaswag.HellaSwag,
jon-tow's avatar
jon-tow committed
130
    "swag": swag.SWAG,
Jon Tow's avatar
Jon Tow committed
131
    "openbookqa": openbookqa.OpenBookQA,
Leo Gao's avatar
Leo Gao committed
132
    "squad2": squad.SQuAD2,
Jon Tow's avatar
Jon Tow committed
133
    "race": race.RACE,
134
    # "naturalqs": naturalqs.NaturalQs, # not implemented yet
Fabrizio Milo's avatar
Fabrizio Milo committed
135
    "headqa": headqa.HeadQAEsDeprecated,  # for backwards compat - headqa used to default to es
136
137
    "headqa_es": headqa.HeadQAEs,
    "headqa_en": headqa.HeadQAEn,
138
    "mathqa": mathqa.MathQA,
139
    "webqs": webqs.WebQs,
140
    "wsc273": wsc273.WinogradSchemaChallenge273,
141
    "winogrande": winogrande.Winogrande,
Jonathan Tow's avatar
Jonathan Tow committed
142
143
144
    "anli_r1": anli.ANLIRound1,
    "anli_r2": anli.ANLIRound2,
    "anli_r3": anli.ANLIRound3,
Leo Gao's avatar
Leo Gao committed
145
146
147
148
149
150
    "ethics_cm": hendrycks_ethics.EthicsCM,
    "ethics_deontology": hendrycks_ethics.EthicsDeontology,
    "ethics_justice": hendrycks_ethics.EthicsJustice,
    "ethics_utilitarianism_original": hendrycks_ethics.EthicsUtilitarianismOriginal,
    "ethics_utilitarianism": hendrycks_ethics.EthicsUtilitarianism,
    "ethics_virtue": hendrycks_ethics.EthicsVirtue,
Fabrizio Milo's avatar
Fabrizio Milo committed
151
152
    "truthfulqa_mc": truthfulqa.TruthfulQAMultipleChoice,
    "truthfulqa_gen": truthfulqa.TruthfulQAGeneration,
Jonathan Tow's avatar
Jonathan Tow committed
153
154
155
    # dialogue
    "mutual": mutual.MuTual,
    "mutual_plus": mutual.MuTualPlus,
Eric Tang's avatar
Eric Tang committed
156
    # math
Leo Gao's avatar
Leo Gao committed
157
158
159
160
161
162
163
    "math_algebra": hendrycks_math.MathAlgebra,
    "math_counting_and_prob": hendrycks_math.MathCountingAndProbability,
    "math_geometry": hendrycks_math.MathGeometry,
    "math_intermediate_algebra": hendrycks_math.MathIntermediateAlgebra,
    "math_num_theory": hendrycks_math.MathNumberTheory,
    "math_prealgebra": hendrycks_math.MathPrealgebra,
    "math_precalc": hendrycks_math.MathPrecalculus,
rokosbasilisk's avatar
rokosbasilisk committed
164
    "math_asdiv": asdiv.Asdiv,
Jonathan Tow's avatar
Jonathan Tow committed
165
    "gsm8k": gsm8k.GradeSchoolMath8K,
166
167
168
169
170
171
172
173
174
175
176
    # arithmetic
    "arithmetic_2da": arithmetic.Arithmetic2DPlus,
    "arithmetic_2ds": arithmetic.Arithmetic2DMinus,
    "arithmetic_3da": arithmetic.Arithmetic3DPlus,
    "arithmetic_3ds": arithmetic.Arithmetic3DMinus,
    "arithmetic_4da": arithmetic.Arithmetic4DPlus,
    "arithmetic_4ds": arithmetic.Arithmetic4DMinus,
    "arithmetic_5da": arithmetic.Arithmetic5DPlus,
    "arithmetic_5ds": arithmetic.Arithmetic5DMinus,
    "arithmetic_2dm": arithmetic.Arithmetic2DMultiplication,
    "arithmetic_1dc": arithmetic.Arithmetic1DComposite,
&'s avatar
& committed
177
178
    # TODO Perhaps make these groups of tasks
    #   e.g. anli, arithmetic, openai_translations, harness_translations
Andy Zou's avatar
Andy Zou committed
179
    # hendrycksTest (57 tasks)
Andy Zou's avatar
Andy Zou committed
180
    **hendrycks_test.create_all_tasks(),
&'s avatar
...  
& committed
181
    # e.g. wmt14-fr-en
&'s avatar
& committed
182
183
184
    **translation.create_tasks_from_benchmarks(gpt3_translation_benchmarks),
    # chef's selection, mostly wmt20
    **translation.create_tasks_from_benchmarks(selected_translation_benchmarks),
Jon Tow's avatar
Jon Tow committed
185
    # Word Scrambling and Manipulation Tasks
186
187
188
189
190
    "anagrams1": unscramble.Anagrams1,
    "anagrams2": unscramble.Anagrams2,
    "cycle_letters": unscramble.CycleLetters,
    "random_insertion": unscramble.RandomInsertion,
    "reversed_words": unscramble.ReversedWords,
Jason Phang's avatar
Jason Phang committed
191
    # Pile
Leo Gao's avatar
Leo Gao committed
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
    "pile_arxiv": pile.PileArxiv,
    "pile_books3": pile.PileBooks3,
    "pile_bookcorpus2": pile.PileBookCorpus2,
    "pile_dm-mathematics": pile.PileDmMathematics,
    "pile_enron": pile.PileEnron,
    "pile_europarl": pile.PileEuroparl,
    "pile_freelaw": pile.PileFreeLaw,
    "pile_github": pile.PileGithub,
    "pile_gutenberg": pile.PileGutenberg,
    "pile_hackernews": pile.PileHackernews,
    "pile_nih-exporter": pile.PileNIHExporter,
    "pile_opensubtitles": pile.PileOpenSubtitles,
    "pile_openwebtext2": pile.PileOpenWebText2,
    "pile_philpapers": pile.PilePhilPapers,
    "pile_pile-cc": pile.PilePileCc,
    "pile_pubmed-abstracts": pile.PilePubmedAbstracts,
    "pile_pubmed-central": pile.PilePubmedCentral,
    "pile_stackexchange": pile.PileStackExchange,
    "pile_uspto": pile.PileUspto,
    "pile_ubuntu-irc": pile.PileUbuntuIrc,
    "pile_wikipedia": pile.PileWikipedia,
    "pile_youtubesubtitles": pile.PileYoutubeSubtitles,
Igor Ostrovsky's avatar
Igor Ostrovsky committed
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
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
    # BLiMP
    "blimp_adjunct_island": blimp.BlimpAdjunctIsland,
    "blimp_anaphor_gender_agreement": blimp.BlimpAnaphorGenderAgreement,
    "blimp_anaphor_number_agreement": blimp.BlimpAnaphorNumberAgreement,
    "blimp_animate_subject_passive": blimp.BlimpAnimateSubjectPassive,
    "blimp_animate_subject_trans": blimp.BlimpAnimateSubjectTrans,
    "blimp_causative": blimp.BlimpCausative,
    "blimp_complex_NP_island": blimp.BlimpComplex_NPIsland,
    "blimp_coordinate_structure_constraint_complex_left_branch": blimp.BlimpCoordinateStructureConstraintComplexLeftBranch,
    "blimp_coordinate_structure_constraint_object_extraction": blimp.BlimpCoordinateStructureConstraintObjectExtraction,
    "blimp_determiner_noun_agreement_1": blimp.BlimpDeterminerNounAgreement_1,
    "blimp_determiner_noun_agreement_2": blimp.BlimpDeterminerNounAgreement_2,
    "blimp_determiner_noun_agreement_irregular_1": blimp.BlimpDeterminerNounAgreementIrregular_1,
    "blimp_determiner_noun_agreement_irregular_2": blimp.BlimpDeterminerNounAgreementIrregular_2,
    "blimp_determiner_noun_agreement_with_adj_2": blimp.BlimpDeterminerNounAgreementWithAdj_2,
    "blimp_determiner_noun_agreement_with_adj_irregular_1": blimp.BlimpDeterminerNounAgreementWithAdjIrregular_1,
    "blimp_determiner_noun_agreement_with_adj_irregular_2": blimp.BlimpDeterminerNounAgreementWithAdjIrregular_2,
    "blimp_determiner_noun_agreement_with_adjective_1": blimp.BlimpDeterminerNounAgreementWithAdjective_1,
    "blimp_distractor_agreement_relational_noun": blimp.BlimpDistractorAgreementRelationalNoun,
    "blimp_distractor_agreement_relative_clause": blimp.BlimpDistractorAgreementRelativeClause,
    "blimp_drop_argument": blimp.BlimpDropArgument,
    "blimp_ellipsis_n_bar_1": blimp.BlimpEllipsisNBar_1,
    "blimp_ellipsis_n_bar_2": blimp.BlimpEllipsisNBar_2,
    "blimp_existential_there_object_raising": blimp.BlimpExistentialThereObjectRaising,
    "blimp_existential_there_quantifiers_1": blimp.BlimpExistentialThereQuantifiers_1,
    "blimp_existential_there_quantifiers_2": blimp.BlimpExistentialThereQuantifiers_2,
    "blimp_existential_there_subject_raising": blimp.BlimpExistentialThereSubjectRaising,
    "blimp_expletive_it_object_raising": blimp.BlimpExpletiveItObjectRaising,
    "blimp_inchoative": blimp.BlimpInchoative,
    "blimp_intransitive": blimp.BlimpIntransitive,
    "blimp_irregular_past_participle_adjectives": blimp.BlimpIrregularPastParticipleAdjectives,
    "blimp_irregular_past_participle_verbs": blimp.BlimpIrregularPastParticipleVerbs,
    "blimp_irregular_plural_subject_verb_agreement_1": blimp.BlimpIrregularPluralSubjectVerbAgreement_1,
    "blimp_irregular_plural_subject_verb_agreement_2": blimp.BlimpIrregularPluralSubjectVerbAgreement_2,
    "blimp_left_branch_island_echo_question": blimp.BlimpLeftBranchIslandEchoQuestion,
    "blimp_left_branch_island_simple_question": blimp.BlimpLeftBranchIslandSimpleQuestion,
    "blimp_matrix_question_npi_licensor_present": blimp.BlimpMatrixQuestionNpiLicensorPresent,
    "blimp_npi_present_1": blimp.BlimpNpiPresent_1,
    "blimp_npi_present_2": blimp.BlimpNpiPresent_2,
    "blimp_only_npi_licensor_present": blimp.BlimpOnlyNpiLicensorPresent,
    "blimp_only_npi_scope": blimp.BlimpOnlyNpiScope,
    "blimp_passive_1": blimp.BlimpPassive_1,
    "blimp_passive_2": blimp.BlimpPassive_2,
    "blimp_principle_A_c_command": blimp.BlimpPrinciple_ACCommand,
    "blimp_principle_A_case_1": blimp.BlimpPrinciple_ACase_1,
    "blimp_principle_A_case_2": blimp.BlimpPrinciple_ACase_2,
    "blimp_principle_A_domain_1": blimp.BlimpPrinciple_ADomain_1,
    "blimp_principle_A_domain_2": blimp.BlimpPrinciple_ADomain_2,
    "blimp_principle_A_domain_3": blimp.BlimpPrinciple_ADomain_3,
    "blimp_principle_A_reconstruction": blimp.BlimpPrinciple_AReconstruction,
    "blimp_regular_plural_subject_verb_agreement_1": blimp.BlimpRegularPluralSubjectVerbAgreement_1,
    "blimp_regular_plural_subject_verb_agreement_2": blimp.BlimpRegularPluralSubjectVerbAgreement_2,
    "blimp_sentential_negation_npi_licensor_present": blimp.BlimpSententialNegationNpiLicensorPresent,
    "blimp_sentential_negation_npi_scope": blimp.BlimpSententialNegationNpiScope,
    "blimp_sentential_subject_island": blimp.BlimpSententialSubjectIsland,
    "blimp_superlative_quantifiers_1": blimp.BlimpSuperlativeQuantifiers_1,
    "blimp_superlative_quantifiers_2": blimp.BlimpSuperlativeQuantifiers_2,
    "blimp_tough_vs_raising_1": blimp.BlimpToughVsRaising_1,
    "blimp_tough_vs_raising_2": blimp.BlimpToughVsRaising_2,
    "blimp_transitive": blimp.BlimpTransitive,
    "blimp_wh_island": blimp.BlimpWhIsland,
    "blimp_wh_questions_object_gap": blimp.BlimpWhQuestionsObjectGap,
    "blimp_wh_questions_subject_gap": blimp.BlimpWhQuestionsSubjectGap,
    "blimp_wh_questions_subject_gap_long_distance": blimp.BlimpWhQuestionsSubjectGapLongDistance,
    "blimp_wh_vs_that_no_gap": blimp.BlimpWhVsThatNoGap,
    "blimp_wh_vs_that_no_gap_long_distance": blimp.BlimpWhVsThatNoGapLongDistance,
    "blimp_wh_vs_that_with_gap": blimp.BlimpWhVsThatWithGap,
    "blimp_wh_vs_that_with_gap_long_distance": blimp.BlimpWhVsThatWithGapLongDistance,
Jonathan Tow's avatar
Jonathan Tow committed
282
283
284
285
    # Requires manual download of data.
    # "storycloze_2016": storycloze.StoryCloze2016,
    # "storycloze_2018": storycloze.StoryCloze2018,
    # "sat": sat.SATAnalogies,
Jason Phang's avatar
Jason Phang committed
286
}
Jason Phang's avatar
gpt3  
Jason Phang committed
287
288


Jason Phang's avatar
Jason Phang committed
289
ALL_TASKS = sorted(list(TASK_REGISTRY))
Jason Phang's avatar
Jason Phang committed
290
291


Jason Phang's avatar
cleanup  
Jason Phang committed
292
def get_task(task_name):
&'s avatar
& committed
293
294
    try:
        return TASK_REGISTRY[task_name]
295
    except KeyError:
&'s avatar
& committed
296
297
298
        print("Available tasks:")
        pprint(TASK_REGISTRY)
        raise KeyError(f"Missing task {task_name}")
Jason Phang's avatar
cleanup  
Jason Phang committed
299
300


301
302
303
304
def get_task_name_from_object(task_object):
    for name, class_ in TASK_REGISTRY.items():
        if class_ is task_object:
            return name
Fabrizio Milo's avatar
Fabrizio Milo committed
305

306
    # this gives a mechanism for non-registered tasks to have a custom name anyways when reporting
Fabrizio Milo's avatar
Fabrizio Milo committed
307
308
309
310
311
    return (
        task_object.EVAL_HARNESS_NAME
        if hasattr(task_object, "EVAL_HARNESS_NAME")
        else type(task_object).__name__
    )
312
313
314


def get_task_dict(task_name_list: List[Union[str, lm_eval.base.Task]]):
315
    task_name_dict = {
Jason Phang's avatar
cleanup  
Jason Phang committed
316
        task_name: get_task(task_name)()
Fabrizio Milo's avatar
Fabrizio Milo committed
317
318
        for task_name in task_name_list
        if isinstance(task_name, str)
319
320
    }
    task_name_from_object_dict = {
321
        get_task_name_from_object(task_object): task_object
Fabrizio Milo's avatar
Fabrizio Milo committed
322
323
        for task_object in task_name_list
        if not isinstance(task_object, str)
Jason Phang's avatar
cleanup  
Jason Phang committed
324
    }
325
326
    assert set(task_name_dict.keys()).isdisjoint(set(task_name_from_object_dict.keys()))
    return {**task_name_dict, **task_name_from_object_dict}