def eval_net():
'''eval net'''
if config.dataset == 'MR':
instance = MovieReview(root_dir=config.data_path, maxlen=config.word_len, split=0.9)
elif config.dataset == 'SUBJ':
instance = Subjectivity(root_dir=config.data_path, maxlen=config.word_len, split=0.9)
elif config.dataset == 'SST2':
instance = SST2(root_dir=config.data_path, maxlen=config.word_len, split=0.9)
device_target = config.device_target
context.set_context(mode=context.GRAPH_MODE, device_target=config.device_target)
if device_target == "Ascend":
context.set_context(device_id=get_device_id())
dataset = instance.create_test_dataset(batch_size=config.batch_size)
loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
net = TextCNN(vocab_len=instance.get_dict_len(), word_len=config.word_len,
num_classes=config.num_classes, vec_length=config.vec_length)
opt = nn.Adam(filter(lambda x: x.requires_grad, net.get_parameters()), learning_rate=0.001,
weight_decay=float(config.weight_decay))
param_dict = load_checkpoint(config.checkpoint_file_path)
print("load checkpoint from [{}].".format(config.checkpoint_file_path))
load_param_into_net(net, param_dict)
net.set_train(False)
model = Model(net, loss_fn=loss, optimizer=opt, metrics={'acc': Accuracy()})
acc = model.eval(dataset)
print("accuracy: ", acc)
help="device target")
parser.add_argument('--dataset', type=str, default='MR', choices=['MR', 'SUBJ', 'SST2'],
help='dataset name.')
args = parser.parse_args()
context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target, device_id=args.device_id)
if __name__ == '__main__':
if args.dataset == 'MR':
cfg = cfg_mr
instance = MovieReview(root_dir=cfg.data_path, maxlen=cfg.word_len, split=0.9)
elif args.dataset == 'SUBJ':
cfg = cfg_subj
instance = Subjectivity(root_dir=cfg.data_path, maxlen=cfg.word_len, split=0.9)
elif args.dataset == 'SST2':
cfg = cfg_sst2
instance = SST2(root_dir=cfg.data_path, maxlen=cfg.word_len, split=0.9)
else:
raise ValueError("dataset is not support.")
net = TextCNN(vocab_len=instance.get_dict_len(), word_len=cfg.word_len,
num_classes=cfg.num_classes, vec_length=cfg.vec_length)
param_dict = load_checkpoint(args.ckpt_file)
load_param_into_net(net, param_dict)
input_arr = Tensor(np.ones([cfg.batch_size, cfg.word_len], np.int32))
export(net, input_arr, file_name=args.file_name, file_format=args.file_format)
def train_net():
'''train net'''
# set context
context.set_context(mode=context.GRAPH_MODE,
device_target=config.device_target)
context.set_context(device_id=get_device_id())
if config.dataset == 'MR':
instance = MovieReview(root_dir=config.data_path,
maxlen=config.word_len,
split=0.9)
elif config.dataset == 'SUBJ':
instance = Subjectivity(root_dir=config.data_path,
maxlen=config.word_len,
split=0.9)
elif config.dataset == 'SST2':
instance = SST2(root_dir=config.data_path,
maxlen=config.word_len,
split=0.9)
dataset = instance.create_train_dataset(batch_size=config.batch_size,
epoch_size=config.epoch_size)
batch_num = dataset.get_dataset_size()
base_lr = float(config.base_lr)
learning_rate = []
warm_up = [
base_lr / math.floor(config.epoch_size / 5) * (i + 1)
for _ in range(batch_num)
for i in range(math.floor(config.epoch_size / 5))
]
shrink = [
base_lr / (16 * (i + 1)) for _ in range(batch_num)
for i in range(math.floor(config.epoch_size * 3 / 5))
]
normal_run = [
base_lr for _ in range(batch_num)
for i in range(config.epoch_size - math.floor(config.epoch_size / 5) -
math.floor(config.epoch_size * 2 / 5))
]
learning_rate = learning_rate + warm_up + normal_run + shrink
net = TextCNN(vocab_len=instance.get_dict_len(),
word_len=config.word_len,
num_classes=config.num_classes,
vec_length=config.vec_length)
# Continue training if set pre_trained to be True
if config.pre_trained:
param_dict = load_checkpoint(config.checkpoint_path)
load_param_into_net(net, param_dict)
opt = nn.Adam(filter(lambda x: x.requires_grad, net.get_parameters()), \
learning_rate=learning_rate, weight_decay=float(config.weight_decay))
loss = SoftmaxCrossEntropyExpand(sparse=True)
model = Model(net,
loss_fn=loss,
optimizer=opt,
metrics={'acc': Accuracy()})
config_ck = CheckpointConfig(
save_checkpoint_steps=int(config.epoch_size * batch_num / 2),
keep_checkpoint_max=config.keep_checkpoint_max)
time_cb = TimeMonitor(data_size=batch_num)
ckpt_save_dir = os.path.join(config.output_path, config.checkpoint_path)
ckpoint_cb = ModelCheckpoint(prefix="train_textcnn",
directory=ckpt_save_dir,
config=config_ck)
loss_cb = LossMonitor()
model.train(config.epoch_size,
dataset,
callbacks=[time_cb, ckpoint_cb, loss_cb])
print("train success")