__init__.py 12.2 KB
Newer Older
&'s avatar
& committed
1
2
from pprint import pprint

&'s avatar
& committed
3
4
import sacrebleu

Jason Phang's avatar
Jason Phang committed
5
6
from . import superglue
from . import glue
Leo Gao's avatar
Leo Gao committed
7
from . import arc
thefazzer's avatar
thefazzer committed
8
from . import coqa
Leo Gao's avatar
Leo Gao committed
9
from . import race
Leo Gao's avatar
Leo Gao committed
10
from . import webqs
Leo Gao's avatar
Leo Gao committed
11
from . import anli
12
from . import wsc273
Charles Foster's avatar
Charles Foster committed
13
from . import winogrande
Charles Foster's avatar
Charles Foster committed
14
from . import quac
Charles Foster's avatar
Charles Foster committed
15
from . import hellaswag
Charles Foster's avatar
Charles Foster committed
16
from . import openbookqa
Charles Foster's avatar
Charles Foster committed
17
from . import squad
18
from . import naturalqs
19
from . import sat
20
from . import arithmetic
Leo Gao's avatar
Leo Gao committed
21
from . import lambada
Jon Tow's avatar
Jon Tow committed
22
from . import race
Leo Gao's avatar
Leo Gao committed
23
from . import piqa
Jonathan Tow's avatar
Jonathan Tow committed
24
from . import prost
Jonathan Tow's avatar
Jonathan Tow committed
25
from . import mc_taco
26
from . import triviaqa
jeffhsu3's avatar
jeffhsu3 committed
27
from . import pubmedqa
28
from . import sciq
29
from . import webqs
30
from . import qa4mre
&'s avatar
& committed
31
from . import translation
32
33
from . import headqa
from . import mathqa
Leo Gao's avatar
Leo Gao committed
34
from . import hendrycks_ethics
Jon Tow's avatar
Jon Tow committed
35
from . import drop
36
from . import unscramble
37
from . import logiqa
Andy Zou's avatar
Andy Zou committed
38
from . import hendrycks_test
Leo Gao's avatar
Leo Gao committed
39
from . import hendrycks_math
40
from . import cbt
41
from . import lambada_cloze
Jason Phang's avatar
Jason Phang committed
42
from . import pile
Leo Gao's avatar
Leo Gao committed
43
from . import wikitext
sdtblck's avatar
sdtblck committed
44
from . import lambada_multilingual
Jonathan Tow's avatar
Jonathan Tow committed
45
from . import mutual
46
from . import truthfulqa
Igor Ostrovsky's avatar
Igor Ostrovsky committed
47
from . import blimp
Jason Phang's avatar
gpt3  
Jason Phang committed
48

&'s avatar
& committed
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
########################################
# Translation tasks
########################################

# 6 total
gpt3_translation_benchmarks = {
    "wmt14": ['en-fr', 'fr-en'],  # French
    "wmt16": ['en-ro', 'ro-en', 'de-en', 'en-de'],  # German, Romanian
}


# 28 total
selected_translation_benchmarks = {
    **gpt3_translation_benchmarks,
    "wmt20": sacrebleu.get_langpairs_for_testset("wmt20"),
    "iwslt17": ['en-ar', 'ar-en']  # Arabic
}

# 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
79
TASK_REGISTRY = {
Jason Phang's avatar
multirc  
Jason Phang committed
80
    # GLUE
Jason Phang's avatar
Jason Phang committed
81
82
    "cola": glue.CoLA,
    "mnli": glue.MNLI,
Jason Phang's avatar
Jason Phang committed
83
    "mnli_mismatched": glue.MNLIMismatched,
Jason Phang's avatar
Jason Phang committed
84
85
86
87
    "mrpc": glue.MRPC,
    "rte": glue.RTE,
    "qnli": glue.QNLI,
    "qqp": glue.QQP,
88
    #"stsb": glue.STSB, # not implemented yet
Jason Phang's avatar
Jason Phang committed
89
90
    "sst": glue.SST,
    "wnli": glue.WNLI,
Jason Phang's avatar
multirc  
Jason Phang committed
91
    # SuperGLUE
Jason Phang's avatar
Jason Phang committed
92
    "boolq": superglue.BoolQ,
thefazzer's avatar
thefazzer committed
93
    "cb": superglue.CommitmentBank,
Jason Phang's avatar
Jason Phang committed
94
    "copa": superglue.Copa,
Jason Phang's avatar
multirc  
Jason Phang committed
95
    "multirc": superglue.MultiRC,
Jason Phang's avatar
Jason Phang committed
96
    "record": superglue.ReCoRD,
Jason Phang's avatar
Jason Phang committed
97
    "wic": superglue.WordsInContext,
Jason Phang's avatar
wsc  
Jason Phang committed
98
    "wsc": superglue.SGWinogradSchemaChallenge,
99
    
Jason Phang's avatar
multirc  
Jason Phang committed
100
    # Order by benchmark/genre?
thefazzer's avatar
thefazzer committed
101
    "coqa": coqa.CoQA,
Jon Tow's avatar
Jon Tow committed
102
    "drop": drop.DROP,
Leo Gao's avatar
Leo Gao committed
103
    "lambada": lambada.LAMBADA,
104
    "lambada_cloze": lambada_cloze.LAMBADA_cloze,
sdtblck's avatar
sdtblck committed
105
    
106
    # multilingual lambada
sdtblck's avatar
sdtblck committed
107
    **lambada_multilingual.construct_tasks(),
sdtblck's avatar
sdtblck committed
108

Leo Gao's avatar
Leo Gao committed
109
    "wikitext": wikitext.WikiText,
Leo Gao's avatar
Leo Gao committed
110
111
    # "cbt-cn": cbt.CBTCN, # disabled pending context length fix
    # "cbt-ne": cbt.CBTNE, # disabled pending context length fix
112

Leo Gao's avatar
Leo Gao committed
113
    "piqa": piqa.PiQA,
Jonathan Tow's avatar
Jonathan Tow committed
114
    "prost": prost.PROST,
Jonathan Tow's avatar
Jonathan Tow committed
115
    "mc_taco": mc_taco.MCTACO,
Leo Gao's avatar
Leo Gao committed
116

117
    # Science related
jeffhsu3's avatar
jeffhsu3 committed
118
    "pubmedqa" : pubmedqa.Pubmed_QA,
119
    "sciq" : sciq.SciQ,
120

121
122
123
    "qa4mre_2011" : qa4mre.QA4MRE_2011,
    "qa4mre_2012" : qa4mre.QA4MRE_2012,
    "qa4mre_2013" : qa4mre.QA4MRE_2013,
jeffhsu3's avatar
jeffhsu3 committed
124

Leo Gao's avatar
Leo Gao committed
125
    "triviaqa": triviaqa.TriviaQA,
Jonathan Tow's avatar
Jonathan Tow committed
126
127
    "arc_easy": arc.ARCEasy,
    "arc_challenge": arc.ARCChallenge,
128
    # "quac": quac.QuAC, # not implemented yet
129
    "logiqa": logiqa.LogiQA,
Leo Gao's avatar
Leo Gao committed
130
    "hellaswag": hellaswag.HellaSwag,
Jon Tow's avatar
Jon Tow committed
131
    "openbookqa": openbookqa.OpenBookQA,
132
    # "sat": sat.SATAnalogies, # not implemented yet
Leo Gao's avatar
Leo Gao committed
133
    "squad2": squad.SQuAD2,
Jon Tow's avatar
Jon Tow committed
134
    "race": race.RACE,
135
    # "naturalqs": naturalqs.NaturalQs, # not implemented yet
136
    "headqa": headqa.HeadQAEsDeprecated, # for backwards compat - headqa used to default to es
137
138
    "headqa_es": headqa.HeadQAEs,
    "headqa_en": headqa.HeadQAEn,
139
    "mathqa": mathqa.MathQA,
140
    "webqs": webqs.WebQs,
141
    "wsc273": wsc273.WinogradSchemaChallenge273,
142
    "winogrande": winogrande.Winogrande,
Jonathan Tow's avatar
Jonathan Tow committed
143
144
145
    "anli_r1": anli.ANLIRound1,
    "anli_r2": anli.ANLIRound2,
    "anli_r3": anli.ANLIRound3,
Muennighoff's avatar
Muennighoff committed
146

Leo Gao's avatar
Leo Gao committed
147
148
149
150
151
152
    "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,
Muennighoff's avatar
Muennighoff committed
153

154
155
     "truthfulqa_mc": truthfulqa.TruthfulQAMultipleChoice,
     "truthfulqa_gen": truthfulqa.TruthfulQAGeneration,
Jonathan Tow's avatar
Jonathan Tow committed
156

Jonathan Tow's avatar
Jonathan Tow committed
157
158
159
160
    # dialogue
    "mutual": mutual.MuTual,
    "mutual_plus": mutual.MuTualPlus,

Eric Tang's avatar
Eric Tang committed
161
    # math
Leo Gao's avatar
Leo Gao committed
162
163
164
165
166
167
168
    "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,
Eric Tang's avatar
Eric Tang committed
169

170
171
172
173
174
175
176
177
178
179
180
    # 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
181
182
    # TODO Perhaps make these groups of tasks
    #   e.g. anli, arithmetic, openai_translations, harness_translations
&'s avatar
...  
& committed
183

Andy Zou's avatar
Andy Zou committed
184
    # hendrycksTest (57 tasks)
Andy Zou's avatar
Andy Zou committed
185
    **hendrycks_test.create_all_tasks(),
Andy Zou's avatar
Andy Zou committed
186

&'s avatar
...  
& committed
187
    # e.g. wmt14-fr-en
&'s avatar
& committed
188
189
190
    **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
191
192

    # Word Scrambling and Manipulation Tasks
193
194
195
196
197
    "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
198
199

    # Pile
Leo Gao's avatar
Leo Gao committed
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
    "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,
sdtblck's avatar
sdtblck committed
222
    
Igor Ostrovsky's avatar
Igor Ostrovsky committed
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
282
283
284
285
286
287
288
289
290
    # 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,
Jason Phang's avatar
Jason Phang committed
291
}
Jason Phang's avatar
gpt3  
Jason Phang committed
292
293


Jason Phang's avatar
Jason Phang committed
294
ALL_TASKS = sorted(list(TASK_REGISTRY))
Jason Phang's avatar
Jason Phang committed
295
296


Jason Phang's avatar
cleanup  
Jason Phang committed
297
def get_task(task_name):
&'s avatar
& committed
298
299
300
301
302
303
    try:
        return TASK_REGISTRY[task_name]
    except KeyError as e:
        print("Available tasks:")
        pprint(TASK_REGISTRY)
        raise KeyError(f"Missing task {task_name}")
Jason Phang's avatar
cleanup  
Jason Phang committed
304
305
306
307
308
309
310


def get_task_dict(task_name_list):
    return {
        task_name: get_task(task_name)()
        for task_name in task_name_list
    }