if __name__ == "__main__":
Log.AppLogger.configure(mkdir=True)
App.add_command(
'decode', decode, {
'gpu':
arg('--gpu',
'-g',
type=int,
default=-1,
help='GPU ID (negative value indicates CPU)'),
'model_file':
arg('--modelfile',
type=str,
required=True,
help='Trained model archive file'),
'save_to':
arg('--out',
type=str,
default=None,
help='Save results to the specified directory'),
'target_file':
arg('--targetfile',
type=str,
required=True,
help='Decoding target data file'),
})
App.run()
if __name__ == "__main__":
App.configure(name='chainer-mnist', logoption='d')
App.add_command(
'train',
train, {
'batch_size':
arg('--batchsize',
'-b',
type=int,
default=100,
help='Number of images in each mini-batch'),
'device':
arg('--device',
type=int,
default=-1,
metavar='ID',
help='Device ID (negative value indicates CPU)'),
'n_epoch':
arg('--epoch',
'-e',
type=int,
default=20,
help='Number of sweeps over the dataset to train'),
'n_units':
arg('--unit', '-u', type=int, default=1000,
help='Number of units'),
},
description="Execute training")
App.run()
App.add_command(
'train', train, {
'batch_size':
arg('--batchsize',
type=int,
default=20,
metavar='NUM',
help='Number of examples in each mini-batch'),
'cache_dir':
arg('--cachedir',
type=str,
default=(App.basedir + '/../cache'),
metavar='DIR',
help='Cache directory'),
'test_file':
arg('--devfile',
type=str,
default=None,
metavar='FILE',
help='Development data file'),
'device':
arg('--device',
type=int,
default=-1,
metavar='ID',
help='Device ID (negative value indicates CPU)'),
'embed_file':
arg('--embedfile',
type=str,
default=None,
metavar='FILE',
help='Pretrained word embedding file'),
'n_epoch':
arg('--epoch',
type=int,
default=20,
metavar='NUM',
help='Number of sweeps over the dataset to train'),
'format':
arg('--format',
type=str,
choices=('tree', 'genia'),
default='tree',
help='Training/Development data format'),
'grad_clip':
arg('--gradclip',
type=float,
default=5.0,
metavar='VALUE',
help='L2 norm threshold of gradient norm'),
'encoder_input':
arg('--inputs',
type=str,
choices=('char', 'postag', 'elmo', 'bert-base', 'bert-large'),
nargs='*',
default=('char', 'postag'),
help='Additional inputs for the encoder'),
'l2_lambda':
arg('--l2',
type=float,
default=0.0,
metavar='VALUE',
help='Strength of L2 regularization'),
'limit':
arg('--limit',
type=int,
default=-1,
metavar='NUM',
help='Limit of the number of training samples'),
'lr':
arg('--lr',
type=float,
default=0.001,
metavar='VALUE',
help='Learning Rate'),
'bert_model':
arg('--bert_model',
type=int,
default=0,
metavar='VALUE',
help='Whether to use BERT model or not'),
'bert_dir':
arg('--bert_dir',
type=str,
default='',
metavar='VALUE',
help='Directory containing bert files'),
'model_config':
arg('--model',
action='store_dict',
metavar='KEY=VALUE',
help='Model configuration'),
'refresh_cache':
arg('--refresh', '-r', action='store_true', help='Refresh cache'),
'save_dir':
arg('--savedir',
type=str,
default=None,
metavar='DIR',
help='Directory to save the model'),
'seed':
arg('--seed',
type=int,
default=None,
metavar='VALUE',
help='Random seed'),
'train_file':
arg('--trainfile',
type=str,
required=True,
metavar='FILE',
help='Training data file'),
})
App.add_command(
'train', train, {
'backend':
arg('--backend',
type=str,
choices=('chainer', 'pytorch'),
default='chainer',
help='Backend framework for computation'),
'batch_size':
arg('--batchsize',
'-b',
type=int,
default=32,
help='Number of examples in each mini-batch'),
'embed_file':
arg('--embedfile',
type=str,
default=None,
help='Pretrained word embedding file'),
'embed_size':
arg('--embedsize',
type=int,
default=100,
help='Size of embeddings'),
'gpu':
arg('--gpu',
'-g',
type=int,
default=-1,
help='GPU ID (negative value indicates CPU)'),
'lr':
arg('--lr', type=float, default=0.002, help='Learning Rate'),
'model_params':
arg('--model',
action='store_dict',
default={},
help='Model hyperparameter'),
'n_epoch':
arg('--epoch',
'-e',
type=int,
default=20,
help='Number of sweeps over the dataset to train'),
'seed':
arg('--seed', type=int, default=None, help='Random seed'),
'save_to':
arg('--out',
type=str,
default=None,
help='Save model to the specified directory'),
'test_file':
arg('--validfile',
type=str,
default=None,
help='validation data file'),
'train_file':
arg('--trainfile',
type=str,
required=True,
help='training data file'),
})
trainer = training.Trainer(optimizer,
model,
loss_func=compute_loss,
accuracy_func=compute_accuracy)
trainer.configure(framework_utils.config)
trainer.fit(train_dataset,
None,
batch_size=batch_size,
epochs=n_epoch,
validation_data=test_dataset,
verbose=App.verbose)
if __name__ == "__main__":
logging.AppLogger.configure(mkdir=True)
App.add_command(
'train', train, {
'train_file': arg('--trainfile', type=str, required=True),
'test_file': arg('--testfile', type=str),
'word_embed_file': arg('--embedfile', type=str),
'n_epoch': arg('--epoch', type=int, default=20),
'batch_size': arg('--batchsize', type=int, default=10),
'lr': arg('--lr', type=float, default=0.01),
'gpu': arg('--gpu', type=int, default=-1),
'seed': arg('--seed', type=int, default=1),
})
chainer.config.debug = False
chainer.config.type_check = False
App.run()
App.add_command(
'train', train, {
'batch_size':
arg('--batchsize',
type=int,
default=5000,
metavar='NUM',
help='Number of tokens in each mini-batch'),
'cache_dir':
arg('--cachedir',
type=str,
default=(App.basedir + '/../cache'),
metavar='DIR',
help='Cache directory'),
'test_file':
arg('--devfile',
type=str,
default=None,
metavar='FILE',
help='Development data file'),
'device':
arg('--device',
type=int,
default=-1,
metavar='ID',
help='Device ID (negative value indicates CPU)'),
'embed_file':
arg('--embedfile',
type=str,
default=None,
metavar='FILE',
help='Pretrained word embedding file'),
'n_epoch':
arg('--epoch',
type=int,
default=300,
metavar='NUM',
help='Number of sweeps over the dataset to train'),
'lr':
arg('--lr',
type=float,
default=2e-3,
metavar='VALUE',
help='Learning rate'),
'model_config':
arg('--model',
action='store_dict',
metavar='KEY=VALUE',
help='Model configuration'),
'refresh_cache':
arg('--refresh', '-r', action='store_true', help='Refresh cache.'),
'save_dir':
arg('--savedir',
type=str,
default=None,
metavar='DIR',
help='Directory to save the model'),
'seed':
arg('--seed',
type=int,
default=None,
metavar='VALUE',
help='Random seed'),
'train_file':
arg('--trainfile',
type=str,
required=True,
metavar='FILE',
help='Training data file.'),
})
App.add_command('train', train, {
'batch_size':
arg('--batchsize', '-b', type=int, default=32,
help='Number of examples in each mini-batch'),
'embed_file':
arg('--embedfile', type=str, default=None,
help='Pretrained word embedding file'),
'embed_size':
arg('--embedsize', type=int, default=100,
help='Size of embeddings'),
'gpu':
arg('--gpu', '-g', type=int, default=-1,
help='GPU ID (negative value indicates CPU)'),
'grad_clip':
arg('--gradclip', type=float, default=5.0,
help='L2 norm threshold of gradient norm'),
'l2_lambda':
arg('--l2', type=float, default=0.0,
help='Strength of L2 regularization'),
'lr':
arg('--lr', type=float, default=0.001,
help='Learning Rate'),
'n_epoch':
arg('--epoch', '-e', type=int, default=20,
help='Number of sweeps over the dataset to train'),
'seed':
arg('--seed', type=int, default=1,
help='Random seed'),
'save_to':
arg('--out', type=str, default=None,
help='Save model to the specified directory'),
'tasks':
arg('--task', type=str, default='tp',
help='Tasks to train: {t: tagging, p: parsing}'),
'test_file':
arg('--validfile', type=str, default=None,
help='validation data file'),
'train_file':
arg('--trainfile', type=str, required=True,
help='training data file'),
})