]
for i in range(len(SENT2VEC)):
s2vsingle[i].load_state(SENT2VEC[i])
s2vsingle[i].set_w2v_path(PATH_TO_W2V)
s2vsingle[i] = s2vsingle[i].cuda()
sent2vec = Sent2Vec(s2vsingle, 'concat')
params_model = {'bsize': 64, 'pool_type': 'mean',
'which_layer': 'all',
'optfile': ELMO_OPTIONS,
'wgtfile': ELMO_WEIGHT}
elmo = ELMo(params_model)
elmo = elmo.cuda()
gensen_1 = GenSenSingle(
model_folder=FOLDER_PATH,
filename_prefix=PREFIX1,
pretrained_emb=PRETRAIN_EMB,
cuda=True
)
gensen_2 = GenSenSingle(
model_folder=FOLDER_PATH,
filename_prefix=PREFIX2,
pretrained_emb=PRETRAIN_EMB,
cuda=True
)
gensen = GenSen(gensen_1, gensen_2)
}
# Set up logger
logging.basicConfig(format='%(asctime)s : %(message)s', level=logging.DEBUG)
if __name__ == "__main__":
# Load InferSent model
params_model = {
'bsize': 64,
'pool_type': 'mean',
'which_layer': 'all',
'optfile': OPT_PATH,
'wgtfile': MODEL_PATH
}
model = ELMo(params_model)
params_senteval['elmo'] = model.cuda()
se = senteval.engine.SE(params_senteval, batcher, prepare)
transfer_tasks = [
'STS12', 'STS13', 'STS14', 'STS15', 'STS16', 'MR', 'CR', 'MPQA',
'SUBJ', 'SST2', 'SST5', 'TREC', 'MRPC', 'SICKEntailment',
'SICKRelatedness', 'STSBenchmark', 'Length', 'WordContent', 'Depth',
'TopConstituents', 'BigramShift', 'Tense', 'SubjNumber', 'ObjNumber',
'OddManOut', 'CoordinationInversion'
]
results_transfer = se.eval(transfer_tasks)
print('--------------------------------------------')
print('MR [Dev:%.1f/Test:%.1f]' %
(results_transfer['MR']['devacc'], results_transfer['MR']['acc']))
print('CR [Dev:%.1f/Test:%.1f]' %