if last or len(generated) > args.max_length:
if len(generated) > 0:
return generated[:args.max_length]
else:
pre_text = generated[-256:]
config = tf.ConfigProto()
if int(args.gpu) >= 0:
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = args.gpu
with tf.Session(config=config,graph=tf.Graph()) as sess:
context = tf.placeholder(tf.int32, [1, None])
output = sample_sequence(
hparams=hparams, length=length,
context=context,
batch_size=1,
temperature=temperature, top_k=top_k, top_p=top_p
)
saver = tf.train.Saver()
ckpt = tf.train.latest_checkpoint(args.model)
saver.restore(sess, ckpt)
if len(args.output_file) > 0:
with open(args.output_file, 'w') as of:
for i in range(args.num_generate):
of.write(generate_one(sess, output)+'\n')
if i < args.num_generate-1:
of.write('========\n')
else:
for i in range(args.num_generate):
n_contexts = hparams.n_ctx - n_prediction
if n_prediction <= 0:
raise ValueError(
'invalid hparams.json or no --max_answer_len argument.')
config = tf.ConfigProto()
if int(args.gpu) >= 0:
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = args.gpu
with tf.Session(config=config) as sess:
context = tf.placeholder(tf.int32, [1, None])
output = sample_sequence(hparams=hparams,
length=hparams.n_ctx,
min_length=min_answer_len,
context=context,
batch_size=1,
temperature=temperature,
top_k=top_k,
top_p=top_p)
saver = tf.train.Saver()
ckpt = tf.train.latest_checkpoint(args.model)
saver.restore(sess, ckpt)
rouge = []
for curDir, dirs, files in os.walk(args.src_dir):
for file in tqdm(files):
if file.endswith(".txt"):
input = os.path.join(curDir, file)
with open(input, 'r', encoding='utf-8') as fp:
raw_text = fp.read()