def main(opt, device_id):
opt = training_opt_postprocessing(opt, device_id)
init_logger(opt.log_file)
out_file = None
best_test_score, best_ckpt = -10000, None
dummy_parser = argparse.ArgumentParser(description='meta_dev_zs.py')
opts.model_opts(dummy_parser)
dummy_opt = dummy_parser.parse_known_args([])[0]
for i in range(57, opt.meta_iterations * opt.inner_iterations + 57, 57 * 5):
ckpt_path = '{}_epoch_{}.pt'.format(opt.save_model, i)
logger.info('Loading checkpoint from %s' % ckpt_path)
if out_file is None:
out_file = codecs.open(opt.output, 'w+', 'utf-8')
fields, model, model_opt = onmt.model_builder.load_test_model(opt, dummy_opt.__dict__, model_path=ckpt_path)
scorer = onmt.translate.GNMTGlobalScorer(opt.alpha,
opt.beta,
opt.coverage_penalty,
opt.length_penalty)
kwargs = {k: getattr(opt, k)
for k in ["beam_size", "n_best", "max_length", "min_length",
"stepwise_penalty", "block_ngram_repeat",
"ignore_when_blocking", "dump_beam", "report_bleu",
"replace_unk", "gpu", "verbose", "fast", "mask_from"]}
fields['graph'] = torchtext.data.Field(sequential = False)
translator = Translator(model, fields, global_scorer=scorer,
out_file=out_file, report_score=False,
copy_attn=model_opt.copy_attn, logger=logger,
log_probs_out_file=None,
**kwargs)
# make translation and save result
all_scores, all_predictions = translator.translate(
src_path='processed_data/meta-dev/' + opt.meta_dev_task + '/src-test.txt',
tgt_path=None,
src_dir=None,
batch_size=opt.translate_batch_size,
attn_debug=False)
# dump predictions
f = open('experiments/meta_dev/' + opt.meta_dev_task + '/test_predictions.csv', 'w', encoding='utf-8')
f.write('smiles,property\n')
for n_best_mols in all_predictions:
for mol in n_best_mols:
f.write(mol.replace(' ', '')+',0\n')
f.close()
# call chemprop to get scores
test_path = '\"' + 'experiments/meta_dev/' + opt.meta_dev_task + '/test_predictions.csv' + '\"'
checkpoint_path = '\"' + 'scorer_ckpts/' + opt.meta_dev_task + '/model.pt' + '\"'
preds_path = '\"' + 'experiments/meta_dev/' + opt.meta_dev_task + '/test_scores.csv' + '\"'
# in case of all mols are invalid (will produce not output file by chemprop)
# the predictions are copied into score file
cmd = 'cp {} {}'.format(test_path, preds_path)
result = os.popen(cmd)
result.close()
cmd = 'python chemprop/predict.py --test_path {} --checkpoint_path {} --preds_path {} --num_workers 0'.format(test_path, checkpoint_path, preds_path)
scorer_result = os.popen(cmd)
# logger.info('{}'.format('\n'.join(scorer_result.readlines())))
scorer_result.close()
# read score file and get score
score = read_score_csv('experiments/meta_dev/' + opt.meta_dev_task + '/test_scores.csv')
assert len(score) % opt.beam_size == 0
# report if it is the best on test
test_metrics = calculate_metrics(opt.meta_dev_task, 'dev', 'test', score)
logger.info('test metrics: ' + str(test_metrics))
test_score = test_metrics['success_rate']
if test_score > best_test_score:
best_ckpt = ckpt_path
logger.info('New best test success rate: {:.4f} by {}'.format(test_score, ckpt_path))
best_test_score = test_score
else:
logger.info('test success rate: {:.4f} by {}'.format(test_score, ckpt_path))
del model_opt
del fields
del model
gc.collect()
def main(opt, device_id):
opt = training_opt_postprocessing(opt, device_id)
init_logger(opt.log_file)
out_file = None
best_test_score, best_ckpt = -10000, None
dummy_parser = argparse.ArgumentParser(description='all_dev.py')
opts.model_opts(dummy_parser)
dummy_opt = dummy_parser.parse_known_args([])[0]
for i in range(0, opt.train_epochs, 10):
ckpt_path = '{}_epoch_{}.pt'.format(opt.save_model, i + 1)
logger.info('Loading checkpoint from %s' % ckpt_path)
checkpoint = torch.load(ckpt_path,
map_location=lambda storage, loc: storage)
model_opt = checkpoint['opt']
fields = load_fields_from_vocab(checkpoint['vocab'], data_type="text")
# Build model.
model = build_model(model_opt, opt, fields, checkpoint)
assert opt.train_from == '' # do not load optimizer state
optim = build_optim(model, opt, checkpoint)
# Build model saver, no need to create task dir for dev
if not os.path.exists('experiments/all_dev'):
os.mkdir('experiments/all_dev')
os.mkdir('experiments/all_dev/' + opt.meta_dev_task)
elif not os.path.exists('experiments/all_dev/' + opt.meta_dev_task):
os.mkdir('experiments/all_dev/' + opt.meta_dev_task)
model_saver = build_model_saver(
model_opt, 'experiments/all_dev/' + opt.meta_dev_task + '/model',
opt, model, fields, optim)
trainer = build_trainer(opt,
device_id,
model,
fields,
optim,
"text",
model_saver=model_saver)
train_iter = list(
build_dataset_iter(lazily_load_dataset("train", opt), fields, opt))
# do training on trainset of meta-dev task
trainer.train(train_iter, opt.inner_iterations)
# do evaluation on devset of meta-dev task
best_dev_score, best_model_path = -10000, None
for model_path in os.listdir('experiments/all_dev/' +
opt.meta_dev_task):
if model_path.find('.pt') == -1:
continue
if out_file is None:
out_file = codecs.open(opt.output, 'w+', 'utf-8')
fields, model, model_opt = onmt.model_builder.load_test_model(
opt,
dummy_opt.__dict__,
model_path='experiments/all_dev/' + opt.meta_dev_task + '/' +
model_path)
scorer = onmt.translate.GNMTGlobalScorer(opt.alpha, opt.beta,
opt.coverage_penalty,
opt.length_penalty)
kwargs = {
k: getattr(opt, k)
for k in [
"beam_size", "n_best", "max_length", "min_length",
"stepwise_penalty", "block_ngram_repeat",
"ignore_when_blocking", "dump_beam", "report_bleu",
"replace_unk", "gpu", "verbose", "fast", "mask_from"
]
}
fields['graph'] = torchtext.data.Field(sequential=False)
translator = Translator(model,
fields,
global_scorer=scorer,
out_file=out_file,
report_score=False,
copy_attn=model_opt.copy_attn,
logger=logger,
log_probs_out_file=None,
**kwargs)
# make translation and save result
all_scores, all_predictions = translator.translate(
src_path='processed_data/meta-dev/' + opt.meta_dev_task +
'/src-dev.txt',
tgt_path=None,
src_dir=None,
batch_size=opt.translate_batch_size,
attn_debug=False)
# dump predictions
f = open('experiments/all_dev/' + opt.meta_dev_task +
'/dev_predictions.csv',
'w',
encoding='utf-8')
f.write('smiles,property\n')
for n_best_mols in all_predictions:
for mol in n_best_mols:
f.write(mol.replace(' ', '') + ',0\n')
f.close()
# call chemprop to get scores
test_path = '\"' + 'experiments/all_dev/' + opt.meta_dev_task + '/dev_predictions.csv' + '\"'
checkpoint_path = '\"' + 'scorer_ckpts/' + opt.meta_dev_task + '/model.pt' + '\"'
preds_path = '\"' + 'experiments/all_dev/' + opt.meta_dev_task + '/dev_scores.csv' + '\"'
# in case of all mols are invalid (will produce not output file by chemprop)
# the predictions are copied into score file
cmd = 'cp {} {}'.format(test_path, preds_path)
result = os.popen(cmd)
result.close()
cmd = 'python chemprop/predict.py --test_path {} --checkpoint_path {} --preds_path {} --num_workers 0'.format(
test_path, checkpoint_path, preds_path)
scorer_result = os.popen(cmd)
scorer_result.close()
# read score file and get score
score = read_score_csv('experiments/all_dev/' + opt.meta_dev_task +
'/dev_scores.csv')
assert len(score) % opt.beam_size == 0
# dev_scores = []
# for i in range(0, len(score), opt.beam_size):
# dev_scores.append(sum([x[1] for x in score[i:i+opt.beam_size]]) / opt.beam_size)
# report dev score
dev_metrics = calculate_metrics(opt.meta_dev_task, 'dev', 'dev',
score)
logger.info('dev metrics: ' + str(dev_metrics))
dev_score = dev_metrics['success_rate']
if dev_score > best_dev_score:
logger.info('New best dev success rate: {:.4f} by {}'.format(
dev_score, model_path))
best_model_path = model_path
best_dev_score = dev_score
else:
logger.info('dev success rate: {:.4f} by {}'.format(
dev_score, model_path))
del fields
del model
del model_opt
del scorer
del translator
gc.collect()
assert best_model_path != None
# do testing on testset of meta-dev task
if out_file is None:
out_file = codecs.open(opt.output, 'w+', 'utf-8')
fields, model, model_opt = onmt.model_builder.load_test_model(
opt,
dummy_opt.__dict__,
model_path='experiments/all_dev/' + opt.meta_dev_task + '/' +
best_model_path)
scorer = onmt.translate.GNMTGlobalScorer(opt.alpha, opt.beta,
opt.coverage_penalty,
opt.length_penalty)
kwargs = {
k: getattr(opt, k)
for k in [
"beam_size", "n_best", "max_length", "min_length",
"stepwise_penalty", "block_ngram_repeat",
"ignore_when_blocking", "dump_beam", "report_bleu",
"replace_unk", "gpu", "verbose", "fast", "mask_from"
]
}
fields['graph'] = torchtext.data.Field(sequential=False)
translator = Translator(model,
fields,
global_scorer=scorer,
out_file=out_file,
report_score=False,
copy_attn=model_opt.copy_attn,
logger=logger,
log_probs_out_file=None,
**kwargs)
# make translation and save result
all_scores, all_predictions = translator.translate(
src_path='processed_data/meta-dev/' + opt.meta_dev_task +
'/src-test.txt',
tgt_path=None,
src_dir=None,
batch_size=opt.translate_batch_size,
attn_debug=False)
# dump predictions
f = open('experiments/all_dev/' + opt.meta_dev_task +
'/test_predictions.csv',
'w',
encoding='utf-8')
f.write('smiles,property\n')
for n_best_mols in all_predictions:
for mol in n_best_mols:
f.write(mol.replace(' ', '') + ',0\n')
f.close()
# call chemprop to get scores
test_path = '\"' + 'experiments/all_dev/' + opt.meta_dev_task + '/test_predictions.csv' + '\"'
checkpoint_path = '\"' + 'scorer_ckpts/' + opt.meta_dev_task + '/model.pt' + '\"'
preds_path = '\"' + 'experiments/all_dev/' + opt.meta_dev_task + '/test_scores.csv' + '\"'
# in case of all mols are invalid (will produce not output file by chemprop)
# the predictions are copied into score file
cmd = 'cp {} {}'.format(test_path, preds_path)
result = os.popen(cmd)
result.close()
cmd = 'python chemprop/predict.py --test_path {} --checkpoint_path {} --preds_path {} --num_workers 0'.format(
test_path, checkpoint_path, preds_path)
scorer_result = os.popen(cmd)
# logger.info('{}'.format('\n'.join(scorer_result.readlines())))
scorer_result.close()
# read score file and get score
score = read_score_csv('experiments/all_dev/' + opt.meta_dev_task +
'/test_scores.csv')
assert len(score) % opt.beam_size == 0
# test_scores = []
# for i in range(0, len(score), opt.beam_size):
# test_scores.append(sum([x[1] for x in score[i:i+opt.beam_size]]) / opt.beam_size)
# report if it is the best on test
test_metrics = calculate_metrics(opt.meta_dev_task, 'dev', 'test',
score)
logger.info('test metrics: ' + str(test_metrics))
test_score = test_metrics['success_rate']
if test_score > best_test_score:
best_ckpt = ckpt_path
logger.info('New best test success rate: {:.4f} by {}'.format(
test_score, ckpt_path))
best_test_score = test_score
else:
logger.info('test success rate: {:.4f} by {}'.format(
test_score, ckpt_path))
del model_opt
del fields
del checkpoint
del model
del optim
del model_saver
del trainer
gc.collect()