def training_start(self, model, data):
training_start_time = time.time()
logger.info("Start training")
# Print a model summary to make sure everything is ok with it
model_summary = torch_summarize(model)
logger.debug(model_summary)
evaluator = BaseEvaluator(self.config)
logger.debug("Preparing training data")
train_batches = data.prepare_training_data(data.train, self.batch_size)
dev_batches = data.prepare_training_data(data.dev, self.batch_size)
id2word = data.vocab.id2tok
dev_lexicalizations = data.lexicalizations['dev']
dev_multi_ref_fn = '%s.multi-ref' % data.fnames['dev']
self.set_optimizer(model, self.config['optimizer'])
self.set_train_criterion(len(id2word), PAD_ID)
#print(data.dev[0])
#exit()
# Moving the model to GPU, if available
if self.use_cuda:
model = model.cuda()
for epoch_idx in range(1, self.n_epochs + 1):
epoch_start = time.time()
pred_fn = os.path.join(self.model_dir, 'predictions.epoch%d' % epoch_idx)
train_loss = self.train_epoch(epoch_idx, model, train_batches)
dev_loss = self.compute_val_loss(model, dev_batches)
predicted_ids, attention_weights = evaluator.evaluate_model(model, data.dev[0], data.uni_mr['dev'])
predicted_tokens = evaluator.lexicalize_predictions(predicted_ids,
dev_lexicalizations,
id2word)
save_predictions_txt(predicted_tokens, pred_fn)
self.record_loss(train_loss, dev_loss)
if self.evaluate_prediction:
self.run_external_eval(dev_multi_ref_fn, pred_fn)
if self.save_model:
save_model(model, os.path.join(self.model_dir, 'weights.epoch%d' % epoch_idx))
logger.info('Epoch %d/%d: time=%s' % (epoch_idx, self.n_epochs, asMinutes(time.time() - epoch_start)))
self.plot_lcurve()
if self.evaluate_prediction:
score_fname = os.path.join(self.model_dir, 'scores.csv')
scores = self.get_scores_to_save()
save_scores(scores, self.score_file_header, score_fname)
self.plot_training_results()
logger.info('End training time=%s' % (asMinutes(time.time() - training_start_time)))
def run(config_dict):
# Fetch all relevant modules.
data_module = config_dict['data-module']
model_module = config_dict['model-module']
training_module = config_dict['training-module']
evaluation_module = config_dict.get('evaluation-module', None)
mode = config_dict['mode']
# Load the modules
DataClass = importlib.import_module(data_module).component
ModelClass = importlib.import_module(model_module).component
TrainingClass = importlib.import_module(training_module).component
EvaluationClass = importlib.import_module(
evaluation_module).component if evaluation_module else None
model_dirname = make_model_dir(config_dict)
logger = set_logger(config_dict["log_level"],
os.path.join(model_dirname, "log.txt"))
# Setup the data
data = DataClass(config_dict["data_params"])
data.setup()
# Setup the model
fix_seed(config_d['random_seed']) # fix seed generators
model = ModelClass(config_dict["model_params"])
model.setup(
data) # there are some data-specific params => pass data as arg
if mode == "train":
training_params = config_dict['training_params']
trainer = TrainingClass(training_params)
trainer.training_start(model, data)
save_config(config_dict, os.path.join(model_dirname, 'config.json'))
elif mode == "predict":
assert evaluation_module is not None, "No evaluation module -- check config file!"
evaluator = EvaluationClass(config_dict)
model_fname = config_dict["model_fn"]
load_model(model, model_fname)
id2word = data.vocab.id2tok
# predict on dev set
if 'dev' in data.fnames:
logger.info("Predicting on dev data")
predicted_ids, attention_weights = evaluator.evaluate_model(
model, data.dev[0])
data_lexicalizations = data.lexicalizations['dev']
predicted_snts = evaluator.lexicalize_predictions(
predicted_ids, data_lexicalizations, id2word)
save_predictions_txt(predicted_snts,
'%s.devset.predictions.txt' % model_fname)
# predict on test set
if 'test' in data.fnames:
logger.info("Predicting on test data")
predicted_ids, attention_weights = evaluator.evaluate_model(
model, data.test[0])
data_lexicalizations = data.lexicalizations['test']
predicted_snts = evaluator.lexicalize_predictions(
predicted_ids, data_lexicalizations, id2word)
save_predictions_txt(predicted_snts,
'%s.testset.predictions.txt' % model_fname)
else:
logger.warning("Check the 'mode' field in the config file: %s" % mode)
logger.info('DONE')
def run(config_dict):
# Fetch all relevant modules.
data_module = config_dict['data-module']
model_module = config_dict['model-module']
training_module = config_dict['training-module']
evaluation_module = config_dict.get('evaluation-module', None)
mode = config_dict['mode']
# Load the modules
DataClass = importlib.import_module(data_module).component
ModelClass = importlib.import_module(model_module).component
TrainingClass = importlib.import_module(training_module).component
EvaluationClass = importlib.import_module(
evaluation_module).component if evaluation_module else None
model_dirname = make_model_dir(config_dict)
logger = set_logger(config_dict["log_level"],
os.path.join(model_dirname, "log.txt"))
# Setup the data
data = DataClass(config_dict["data_params"])
data.setup()
# Setup the model
fix_seed(config_d['random_seed']) # fix seed generators
model = ModelClass(config_dict["model_params"])
print("build model done")
model.setup(
data) # there are some data-specific params => pass data as arg
print("setup data done")
#print(len(data.lexicalizations['test']))
if mode == "train":
training_params = config_dict['training_params']
trainer = TrainingClass(training_params)
trainer.training_start(model, data)
save_config(config_dict, os.path.join(model_dirname, 'config.json'))
elif mode == "predict":
assert evaluation_module is not None, "No evaluation module -- check config file!"
evaluator = EvaluationClass(config_dict)
model_fname = config_dict["model_fn"]
load_model(model, model_fname)
#print(model.state_dict())
model = model.to('cuda')
id2word = data.vocab.id2tok
beam_size = None #10
alpha = 0.3
#"""
if 'dev' in data.fnames:
logger.info("Predicting on dev data")
print(len(data.uni_mr['dev']), len(data.dev[0]),
len(data.lexicalizations['dev']))
dec_snt_beam, fw_beam = [], []
predicted_ids, fw_beam = evaluator.evaluate_model(
model,
data.dev[0],
data.uni_mr['dev'],
beam_size=beam_size,
alpha=alpha)
data_lexicalizations = data.lexicalizations['dev']
#dec_snt_beam = []
#for _ in range(beam_size):
# predicted_snt = evaluator.lexicalize_predictions(predicted_ids[_], data_lexicalizations, id2word)
# dec_snt_beam.append( predicted_snt )
print(len(predicted_ids), len(data_lexicalizations))
predicted_snts = evaluator.lexicalize_predictions(
predicted_ids, data_lexicalizations, id2word)
#save_predictions_txt(predicted_snts, '%s.devset.predictions.txt_incre_0.7' % model_fname)
save_predictions_txt(
predicted_snts,
'%s.devset.predictions.txt_incre_%.1f' % (model_fname, alpha))
# for beam_idx in range(beam_size):
# predicted_ids, attention_weights = evaluator.evaluate_model(model, data.dev[0], data.uni_mr['dev'])
# data_lexicalizations = data.lexicalizations['dev']
# #print(len(predicted_ids), len(data_lexicalizations))
# predicted_snts = evaluator.lexicalize_predictions(predicted_ids,
# data_lexicalizations,
# id2word)
# fw_prob = [[x.item() for x in x_list ] for x_list in attention_weights]
# #print(len(fw_prob), len(fw_prob[0]))#, fw_prob)
# dec_snt_beam.append( predicted_snts ), fw_beam.append( fw_prob )
# save_predictions_txt(predicted_snts, '%s.devset.predictions.txt' % model_fname)
#save_beam_fw(fw_beam, dec_snt_beam, beam_size, '%s.devset.recs.txt' % model_fname)
exit()
#"""
if 'test' in data.fnames:
logger.info("Predicting on test data")
print(len(data.test[0]))
predicted_ids, attention_weights = evaluator.evaluate_model(
model,
data.test[0],
data.uni_mr['test'],
beam_size=beam_size,
alpha=alpha)
data_lexicalizations = data.lexicalizations['test']
print(len(predicted_ids), len(data_lexicalizations))
predicted_snts = evaluator.lexicalize_predictions(
predicted_ids, data_lexicalizations, id2word)
save_predictions_txt(
predicted_snts,
'%s.testset.predictions.txt_inre_%.1f' % (model_fname, alpha))
else:
logger.warning("Check the 'mode' field in the config file: %s" % mode)
logger.info('DONE')
