def __init__(self, hp):
self.hp = hp
self.token2idx, self.idx2token = _load_vocab(hp.vocab)
self.vocab_len = len(self.token2idx)
self.embeddings = get_token_embeddings(self.vocab_len,
self.hp.d_model,
zero_pad=True)
def __init__(self, model_dir, vocab_file):
"""
:param model_dir: model dir path
:param vocab_file: vocab file path
"""
self.tf = import_tf(0)
self.model_dir = model_dir
self.vocab_file = vocab_file
self.token2idx, self.idx2token = _load_vocab(vocab_file)
hparams = Hparams()
parser = hparams.parser
self.hp = parser.parse_args()
self.model = Transformer(self.hp)
self._add_placeholder()
self._init_graph()
# create a iter of the correct shape and type
xs, ys = iter.get_next()
logging.info('# init data')
training_iter = train_batches.make_one_shot_iterator()
val_iter = eval_batches.make_initializable_iterator()
logging.info("# Load model")
m = Transformer(hp)
# get op
loss, train_op, global_step, train_summaries = m.train(xs, ys)
y_hat, eval_summaries = m.eval(xs, ys)
token2idx, idx2token = _load_vocab(hp.vocab)
bs = BeamSearch(m, hp.beam_size,
list(idx2token.keys())[2],
list(idx2token.keys())[3], idx2token, hp.maxlen2, m.x,
m.decoder_inputs, m.logits)
logging.info("# Session")
saver = tf.train.Saver(max_to_keep=hp.num_epochs)
with tf.Session(config=tf.ConfigProto(allow_soft_placement=True)) as sess:
ckpt = tf.train.latest_checkpoint(hp.logdir)
if ckpt is None:
logging.info("Initializing from scratch")
sess.run(tf.global_variables_initializer())
save_variable_specs(os.path.join(hp.logdir, "specs"))
else:
