def main():
logging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
logger = logging.getLogger(__name__)
args = get_args()
print_args(args)
device, n_gpu = initialization.init_cuda_from_args(args, logger=logger)
initialization.init_seed(args, n_gpu=n_gpu, logger=logger)
initialization.init_output_dir(args)
initialization.save_args(args)
classifier = simple_classifier(n_classes=args.n_classes, n_hidden=args.fc_dim)
classifier = classifier.to(device)
optimizer = SGD(classifier.parameters(), lr=0.001, momentum=0.9)
runner = Runner(classifier=classifier,
optimizer=optimizer,
device=device,
rparams=RunnerParameters(
num_train_epochs=args.num_train_epochs,
train_batch_size=args.train_batch_size,
eval_batch_size=args.eval_batch_size,
))
# dataset
train_dataset = torch.load(os.path.join(args.data_dir, "train.dataset"))
eval_dataset = torch.load(os.path.join(args.data_dir, "dev.dataset"))
if args.mnli:
mm_eval_dataset = torch.load(os.path.join(args.data_dir, "mm_dev.dataset"))
else:
mm_eval_dataset = None
# run training and validation
to_save = runner.run_train_val(
train_dataset=train_dataset,
eval_dataset=eval_dataset,
mm_eval_set=mm_eval_dataset,
)
# save training state to output dir.
torch.save(to_save, os.path.join(args.output_dir, "training.info"))
def main():
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
logger = logging.getLogger(__name__)
args = get_args()
print_args(args)
device, n_gpu = initialization.init_cuda_from_args(args, logger=logger)
initialization.init_seed(args, n_gpu=n_gpu, logger=logger)
initialization.init_train_batch_size(args)
initialization.init_output_dir(args)
initialization.save_args(args)
task = get_task(args.task_name, args.data_dir)
use_cuda = False if args.no_cuda else True
verbose = args.verbose
config = {
'dropout_prob': args.dropout_prob,
'n_classes': args.n_classes,
'fc_dim': args.fc_dim,
'enc_dim': 512,
}
# load model
print("loading Universal Sentence Encoder......")
USE = hub.load("https://tfhub.dev/google/universal-sentence-encoder/4")
classifier = SimpleClassifier(config)
classifier = classifier.cuda() if not args.no_cuda else classifier
# get train examples
train_examples = task.get_train_examples()
# calculate t_total
t_total = initialization.get_opt_train_steps(len(train_examples), args)
# build optimizer.
optimizer = optim.SGD(classifier.parameters(), lr=0.001, momentum=0.9)
# create running parameters
r_params = RunnerParameters(
local_rank=args.local_rank,
n_gpu=n_gpu,
learning_rate=5e-5,
gradient_accumulation_steps=args.gradient_accumulation_steps,
t_total=t_total,
warmup_proportion=args.warmup_proportion,
num_train_epochs=args.num_train_epochs,
train_batch_size=args.train_batch_size,
eval_batch_size=args.eval_batch_size,
verbose=verbose)
# create runner class for training and evaluation tasks.
runner = GlueTaskClassifierRunner(encoder_model=USE,
classifier_model=classifier,
optimizer=optimizer,
label_list=task.get_labels(),
device=device,
rparams=r_params)
if args.do_train:
runner.run_train_classifier(train_examples)
if args.do_val:
val_examples = task.get_dev_examples()
results = runner.run_val(val_examples,
task_name=task.name,
verbose=verbose)
df = pd.DataFrame(results["logits"])
df.to_csv(os.path.join(args.output_dir, "val_preds.csv"),
header=False,
index=False)
metrics_str = json.dumps(
{
"loss": results["loss"],
"metrics": results["metrics"]
}, indent=2)
print(metrics_str)
with open(os.path.join(args.output_dir, "val_metrics.json"), "w") as f:
f.write(metrics_str)
# HACK for MNLI-mismatched
if task.name == "mnli":
mm_val_example = MnliMismatchedProcessor().get_dev_examples(
task.data_dir)
mm_results = runner.run_val(mm_val_example,
task_name=task.name,
verbose=verbose)
df = pd.DataFrame(results["logits"])
df.to_csv(os.path.join(args.output_dir, "mm_val_preds.csv"),
header=False,
index=False)
combined_metrics = {}
for k, v in results["metrics"].items():
combined_metrics[k] = v
for k, v in mm_results["metrics"].items():
combined_metrics["mm-" + k] = v
combined_metrics_str = json.dumps(
{
"loss": results["loss"],
"metrics": combined_metrics,
},
indent=2)
print(combined_metrics_str)
with open(os.path.join(args.output_dir, "val_metrics.json"),
"w") as f:
f.write(combined_metrics_str)
def main():
logging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
logger = logging.getLogger(__name__)
args = get_args()
print_args(args)
device, n_gpu = initialization.init_cuda_from_args(args, logger=logger)
initialization.init_seed(args, n_gpu=n_gpu, logger=logger)
initialization.init_output_dir(args)
initialization.save_args(args)
classifier = simple_classifier(n_classes=args.n_classes, n_hidden=args.fc_dim)
classifier = classifier.to(device)
optimizer = SGD(classifier.parameters(), lr=0.001, momentum=0.9)
runner = InfersentRunner(classifier=classifier,
optimizer=optimizer,
device=device,
rparams=RunnerParameters(
num_train_epochs=args.num_train_epochs,
train_batch_size=args.train_batch_size,
eval_batch_size=args.eval_batch_size,
))
# dataset
train_datasets = []
for i in range(TRAIN_SET_NUM_MAP[args.task_name]):
train_dataset = torch.load(os.path.join(args.data_dir, "train-{}.dataset".format(i)))
train_datasets.append(train_dataset)
eval_dataset = torch.load(os.path.join(args.data_dir, "dev-0.dataset"))
if args.mnli:
mm_eval_dataset = torch.load(os.path.join(args.data_dir, "mm_dev-0.dataset"))
else:
mm_eval_dataset = None
# run training and validation
eval_info, state_dicts = runner.run_train_val_with_state_dict_returned(
train_dataset=train_datasets,
eval_dataset=eval_dataset,
mm_eval_set=mm_eval_dataset,
)
# save training state to output dir.
torch.save(eval_info, os.path.join(args.output_dir, "training.info"))
# find highest validation results, load model state dict, and then run prediction @ test set.
val_acc = []
mm_val_acc = []
if args.mnli:
for item in eval_info:
val_acc.append(item[0]['accuracy'])
mm_val_acc.append(item[1]['accuracy'])
idx = val_acc.index(max(val_acc))
print("highest accuracy on validation is: {}, index = {}. \n"
"mis-matched is: {} \n"
"Load state dicts and run testing...".format(val_acc[idx], idx, mm_val_acc[idx]))
else:
for item in eval_info:
val_acc.append(item['accuracy'])
idx = val_acc.index(max(val_acc))
print("highest accuracy on validation is: {}, index = {}. \n"
"Load state dicts and run testing...".format(val_acc[idx], idx))
torch.save(state_dicts[idx], os.path.join(args.output_dir, "state.p"))
test_datasets = []
for i in range(TEST_SET_NUM_MAP[args.task_name]):
test_dataset = torch.load(os.path.join(args.data_dir, "test-{}.dataset".format(i)))
test_datasets.append(test_dataset)
runner.classifier.load_state_dict(torch.load(os.path.join(args.output_dir, "state.p")))
logits = runner.run_test(test_datasets)
df = pd.DataFrame(logits)
df.to_csv(os.path.join(args.output_dir, "test_preds.csv"), header=False, index=False)
# HACK for MNLI-mismatched
if args.mnli:
mm_test_dataset = torch.load(os.path.join(args.data_dir, "mm_test-0.dataset"))
logits = runner.run_test([mm_test_dataset])
df = pd.DataFrame(logits)
df.to_csv(os.path.join(args.output_dir, "mm_test_preds.csv"), header=False, index=False)
def main():
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
logger = logging.getLogger(__name__)
args = get_args()
print_args(args)
device, n_gpu = initialization.init_cuda_from_args(args, logger=logger)
initialization.init_seed(args, n_gpu=n_gpu, logger=logger)
initialization.init_train_batch_size(args)
initialization.init_output_dir(args)
initialization.save_args(args)
task = get_task(args.task_name, args.data_dir)
use_cuda = False if args.no_cuda else True
verbose = args.verbose
# model config
config = {
'word_emb_dim': args.word_emb_dim,
'enc_lstm_dim': args.enc_lstm_dim,
'n_enc_layers': args.n_enc_layers,
'dpout_model': args.dpout_model,
'dpout_fc': args.dpout_fc,
'fc_dim': args.fc_dim,
'bsize': args.batch_size,
'n_classes': args.n_classes,
'pool_type': args.pool_type,
'nonlinear_fc': args.nonlinear_fc,
'use_cuda': use_cuda,
'version': args.model_version,
'dropout_prob': args.dropout_prob,
}
# load model
if verbose:
print('loading model...')
model = InferSent(config)
model.load_state_dict(torch.load(args.model_path))
model = model.cuda() if not args.no_cuda else model
model.set_w2v_path(args.word_emb_path)
model.build_vocab_k_words(K=args.k_freq_words, verbose=verbose)
# load classifier
classifier = SimpleClassifier(config)
classifier = classifier.cuda() if not args.no_cuda else classifier
# get train examples
train_examples = task.get_train_examples()
# calculate t_total
t_total = initialization.get_opt_train_steps(len(train_examples), args)
# build optimizer.
optimizer = optim.SGD(classifier.parameters(), lr=0.001, momentum=0.9)
# create running parameters
r_params = RunnerParameters(
local_rank=args.local_rank,
n_gpu=n_gpu,
learning_rate=5e-5,
gradient_accumulation_steps=args.gradient_accumulation_steps,
t_total=t_total,
warmup_proportion=args.warmup_proportion,
num_train_epochs=args.num_train_epochs,
train_batch_size=args.train_batch_size,
eval_batch_size=args.eval_batch_size,
verbose=verbose)
# create runner class for training and evaluation tasks.
runner = GlueTaskClassifierRunner(encoder_model=model,
classifier_model=classifier,
optimizer=optimizer,
label_list=task.get_labels(),
device=device,
rparams=r_params)
if args.do_train:
runner.run_train_classifier(train_examples)
if args.do_val:
val_examples = task.get_dev_examples()
results = runner.run_val(val_examples,
task_name=task.name,
verbose=verbose)
df = pd.DataFrame(results["logits"])
df.to_csv(os.path.join(args.output_dir, "val_preds.csv"),
header=False,
index=False)
metrics_str = json.dumps(
{
"loss": results["loss"],
"metrics": results["metrics"]
}, indent=2)
print(metrics_str)
with open(os.path.join(args.output_dir, "val_metrics.json"), "w") as f:
f.write(metrics_str)
# HACK for MNLI-mismatched
if task.name == "mnli":
mm_val_example = MnliMismatchedProcessor().get_dev_examples(
task.data_dir)
mm_results = runner.run_val(mm_val_example,
task_name=task.name,
verbose=verbose)
df = pd.DataFrame(results["logits"])
df.to_csv(os.path.join(args.output_dir, "mm_val_preds.csv"),
header=False,
index=False)
combined_metrics = {}
for k, v in results["metrics"].items():
combined_metrics[k] = v
for k, v in mm_results["metrics"].items():
combined_metrics["mm-" + k] = v
combined_metrics_str = json.dumps(
{
"loss": results["loss"],
"metrics": combined_metrics,
},
indent=2)
print(combined_metrics_str)
with open(os.path.join(args.output_dir, "val_metrics.json"),
"w") as f:
f.write(combined_metrics_str)