def main():
# Load vocab.
vocab = Vocab.from_json(args.vocab_path)
label_map = {0: 'negative', 1: 'positive'}
# Constructs the newtork.
network = args.network.lower()
vocab_size = len(vocab)
num_classes = len(label_map)
pad_token_id = vocab.to_indices('[PAD]')
if network == 'bow':
model = BoWModel(vocab_size, num_classes, padding_idx=pad_token_id)
elif network == 'bigru':
model = GRUModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'bilstm':
model = LSTMModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'bilstm_attn':
lstm_hidden_size = 196
attention = SelfInteractiveAttention(hidden_size=2 * lstm_hidden_size)
model = BiLSTMAttentionModel(attention_layer=attention,
vocab_size=vocab_size,
lstm_hidden_size=lstm_hidden_size,
num_classes=num_classes,
padding_idx=pad_token_id)
elif network == 'birnn':
model = RNNModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'cnn':
model = CNNModel(vocab_size, num_classes, padding_idx=pad_token_id)
elif network == 'gru':
model = GRUModel(vocab_size,
num_classes,
direction='forward',
padding_idx=pad_token_id,
pooling_type='max')
elif network == 'lstm':
model = LSTMModel(vocab_size,
num_classes,
direction='forward',
padding_idx=pad_token_id,
pooling_type='max')
elif network == 'rnn':
model = RNNModel(vocab_size,
num_classes,
direction='forward',
padding_idx=pad_token_id,
pooling_type='max')
else:
raise ValueError(
"Unknown network: %s, it must be one of bow, lstm, bilstm, cnn, gru, bigru, rnn, birnn and bilstm_attn."
% network)
# Load model parameters.
state_dict = paddle.load(args.params_path)
model.set_dict(state_dict)
model.eval()
inputs = [paddle.static.InputSpec(shape=[None, None], dtype="int64")]
# Convert to static graph with specific input description
if args.network in [
"lstm", "bilstm", "gru", "bigru", "rnn", "birnn", "bilstm_attn"
]:
inputs.append(paddle.static.InputSpec(shape=[None],
dtype="int64")) # seq_len
model = paddle.jit.to_static(model, input_spec=inputs)
# Save in static graph model.
paddle.jit.save(model, args.output_path)
elif network == 'bigru':
model = GRUModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'bilstm':
model = LSTMModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'bilstm_attn':
lstm_hidden_size = 196
attention = SelfInteractiveAttention(hidden_size=2 * stm_hidden_size)
model = BiLSTMAttentionModel(attention_layer=attention,
vocab_size=vocab_size,
lstm_hidden_size=lstm_hidden_size,
num_classes=num_classes,
padding_idx=pad_token_id)
elif network == 'birnn':
model = RNNModel(vocab_size,
num_classes,
direction='bidirect',
padding_idx=pad_token_id)
elif network == 'cnn':
model = CNNModel(vocab_size, num_classes, padding_idx=pad_token_id)
elif network == 'gru':
model = GRUModel(vocab_size,
num_classes,
direction='forward',
padding_idx=pad_token_id,
pooling_type='max')
vocab = Vocab.load_vocabulary(
args.vocab_path, unk_token='[UNK]', pad_token='[PAD]')
tokenizer = CharTokenizer(vocab, args.language, '../../../punctuations')
# Constructs the newtork.
vocab_size = len(vocab)
num_classes = len(train_ds.label_list)
pad_token_id = 0
pad_value = vocab.token_to_idx.get('[PAD]', 0)
lstm_hidden_size = 196
attention = SelfInteractiveAttention(hidden_size=2 * lstm_hidden_size)
model = BiLSTMAttentionModel(
attention_layer=attention,
vocab_size=vocab_size,
lstm_hidden_size=lstm_hidden_size,
num_classes=num_classes,
padding_idx=pad_token_id)
model = paddle.Model(model)
# Reads data and generates mini-batches.
trans_fn = partial(
convert_example,
tokenizer=tokenizer,
is_test=False,
language=args.language)
batchify_fn = lambda samples, fn=Tuple(
Pad(axis=0, pad_val=pad_value), # input_ids
Stack(dtype="int64"), # seq len