def wmt_test_generation():
src_lang, trg_lang, dataset, test_dataset = get_attributes('test')
script_list = [
WMTGenerationJobScript(outdir, test_dataset, src_lang, trg_lang)
for outdir in get_outdir_list(dataset, 'test')
]
generate_run(script_list, TestGenerationRunScript, TestGenerationSubScript)
def wmt_test_score():
src_lang, trg_lang, dataset, test_dataset = get_attributes('test')
script_list = [
WMTScoreJobScript(outdir, test_dataset, src_lang, trg_lang)
for outdir in get_outdir_list(dataset, 'test')
]
generate_run(script_list, TestScoreRunScript, TestScoreSubScript)
def conll13_single_generation():
outdir_list = get_single_valid_outdir_list('conll')
script_list = [
CoNLL13SingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, CoNLL13SingleGenerationRunScript,
CoNLL13SingleGenerationSubScript)
def test_single_score(dataset, result_table_factory_class,
score_job_script_class, score_run_script_class,
score_sub_script_class):
valid_result_table = result_table_factory_class().make(dataset, 'valid')
outdir_list = get_single_test_outdir_list(dataset, valid_result_table)
script_list = [score_job_script_class(outdir) for outdir in outdir_list]
generate_run(script_list, score_run_script_class, score_sub_script_class)
def ensemble_score(dataset, phase, result_table_factory_class,
score_job_script_class, score_run_script_class,
score_sub_script_class):
valid_result_table = result_table_factory_class().make(dataset, 'valid')
outdir = get_ensemble_outdir(dataset, phase, valid_result_table)
script_list = [score_job_script_class(outdir)]
generate_run(script_list, score_run_script_class, score_sub_script_class)
def fce_valid_single_generation():
outdir_list = get_single_valid_outdir_list('fce')
script_list = [
FCEValidSingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, FCEValidSingleGenerationRunScript,
FCEValidSingleGenerationSubScript)
def jfleg_valid_single_generation():
outdir_list = get_single_valid_outdir_list('jfleg')
script_list = [
JFLEGValidSingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, JFLEGValidSingleGenerationRunScript,
JFLEGValidSingleGenerationSubScript)
def bea19_valid_single_generation():
outdir_list = get_single_valid_outdir_list('bea19')
script_list = [
BEA19ValidSingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, BEA19ValidSingleGenerationRunScript,
BEA19ValidSingleGenerationSubScript)
def wmt_valid_score():
src_lang, trg_lang, dataset, valid_dataset = get_attributes('valid')
script_list = [
WMTScoreJobScript(outdir, valid_dataset, src_lang, trg_lang)
for outdir in get_outdir_list(dataset, 'valid')
]
generate_run(script_list, ValidScoreRunScript, ValidScoreSubScript)
def jfleg_test_single_generation():
valid_result_table = GLEUResultTableFactory().make('jfleg', 'valid')
outdir_list = get_single_test_outdir_list('jfleg', valid_result_table)
script_list = [
JFLEGTestSingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, JFLEGTestSingleGenerationRunScript,
JFLEGTestSingleGenerationSubScript)
def conll14_single_generation():
valid_result_table = M2ResultTableFactory().make('conll', 'valid')
outdir_list = get_single_test_outdir_list('conll', valid_result_table)
script_list = [
CoNLL14SingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, CoNLL14SingleGenerationRunScript,
CoNLL14SingleGenerationSubScript)
def wmt_valid_generation():
src_lang, trg_lang, dataset, valid_dataset = get_attributes('valid')
script_list = [
WMTGenerationJobScript(outdir, valid_dataset, src_lang, trg_lang)
for outdir in get_outdir_list(dataset, 'valid')
]
generate_run(script_list, ValidGenerationRunScript,
ValidGenerationSubScript)
def fce_test_single_generation():
valid_result_table = ErrantResultTableFactory().make('fce', 'valid')
outdir_list = get_single_test_outdir_list('fce', valid_result_table)
script_list = [
FCETestSingleGenerationJobScript(outdir) for outdir in outdir_list
]
generate_run(script_list, FCETestSingleGenerationRunScript,
FCETestSingleGenerationSubScript)
def conll_ensemble_generation():
valid_result_table = M2ResultTableFactory().make('conll', 'valid')
valid_outdir = get_ensemble_outdir('conll', 'valid', valid_result_table)
test_outdir = get_ensemble_outdir('conll', 'test', valid_result_table)
script_list = [
CoNLL13EnsembleGenerationJobScript(valid_outdir),
CoNLL14EnsembleGenerationJobScript(test_outdir)
]
generate_run(script_list, CoNLLEnsembleGenerationRunScript,
CoNLLEnsembleGenerationSubScript)
def bea19_ensemble_generation():
valid_result_table = ErrantResultTableFactory().make('bea19', 'valid')
valid_outdir = get_ensemble_outdir('bea19', 'valid', valid_result_table)
test_outdir = get_ensemble_outdir('bea19', 'test', valid_result_table)
script_list = [
BEA19ValidEnsembleGenerationJobScript(valid_outdir),
BEA19TestEnsembleGenerationJobScript(test_outdir)
]
generate_run(script_list, BEA19EnsembleGenerationRunScript,
BEA19EnsembleGenerationSubScript)
def fce_ensemble_generation():
valid_result_table = ErrantResultTableFactory().make('fce', 'valid')
valid_outdir = get_ensemble_outdir('fce', 'valid', valid_result_table)
test_outdir = get_ensemble_outdir('fce', 'test', valid_result_table)
script_list = [
FCEValidEnsembleGenerationJobScript(valid_outdir),
FCETestEnsembleGenerationJobScript(test_outdir)
]
generate_run(script_list, FCEEnsembleGenerationRunScript,
FCEEnsembleGenerationSubScript)
def jfleg_ensemble_generation():
valid_result_table = GLEUResultTableFactory().make('jfleg', 'valid')
valid_outdir = get_ensemble_outdir('jfleg', 'valid', valid_result_table)
test_outdir = get_ensemble_outdir('jfleg', 'test', valid_result_table)
script_list = [
JFLEGValidEnsembleGenerationJobScript(valid_outdir),
JFLEGTestEnsembleGenerationJobScript(test_outdir)
]
generate_run(script_list, JFLEGEnsembleGenerationRunScript,
JFLEGEnsembleGenerationSubScript)
def ensemble_reranked_score(dataset, phase, score_job_script_class,
score_run_script_class, score_sub_script_class):
arch_list = get_arch_list()
outdir_list = [
get_ensemble_reranking_outdir(dataset, phase, arch)
for arch in arch_list
]
lambda_list = get_lambda_list()
script_list = [
score_job_script_class(outdir, l) for outdir in outdir_list
for l in lambda_list
]
generate_run(script_list, score_run_script_class, score_sub_script_class)
def rtt_merge():
script_list = [RTTMergeJobScript(index) for index in get_rtt_indices()]
generate_run(script_list, RTTMergeRunScript, RTTMergeSubScript)
def tokenize():
config = load_config()
num_iter = config['iter']
script_list = [TokenizeJobScript(index) for index in range(num_iter)]
generate_run(script_list, TokenizeRunScript, TokenizeSubScript)
def ready():
config = load_config()
num_iter = len(config['input_list'])
script_list = [ReadyJobScript(index) for index in range(num_iter)]
generate_run(script_list, ReadyRunScript, ReadySubScript)
def bea19_prepare():
config = load_config()
num_iter = config['iter']
script_list = [BEA19PrepareJobScript(n) for n in range(num_iter)]
generate_run(script_list, PrepareRunScript, BEA19PrepareSubScript)
def valid_single_score(dataset, score_job_script_class, score_run_script_class,
score_sub_script_class):
outdir_list = get_single_valid_outdir_list(dataset)
script_list = [score_job_script_class(outdir) for outdir in outdir_list]
generate_run(script_list, score_run_script_class, score_sub_script_class)
def rtt_split():
script_list = [RTTSplitJobScript(index) for index in get_rtt_indices()]
generate_run(script_list, RTTSplitRunScript, RTTSplitSubScript)
