#
# Copyright 2018-2019 IBM Corp. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from maxfw.model import MAXModelWrapper
from keras.backend import clear_session
from keras import models
import tensorflow as tf
import numpy as np
import json
import logging
from config import DEFAULT_MODEL_PATH, DEFAULT_MODEL_FILE, SEED_TEXT_LEN, MODEL_META_DATA as model_meta
logger = logging.getLogger()
# (Fixed) length of seed text that can serve as input to the generative model
_SEED_TEXT_LEN = 256
class ModelWrapper(MAXModelWrapper):
MODEL_META_DATA = model_meta
"""Model wrapper for Keras models"""
def __init__(self, path=DEFAULT_MODEL_PATH, model_file=DEFAULT_MODEL_FILE):
logger.info('Loading model from: {}...'.format(path))
model_path = '{}/{}'.format(path, model_file)
clear_session()
self.graph = tf.Graph()
with self.graph.as_default():
self.model = models.load_model(model_path)
logger.info('Loaded model: {}'.format(self.model.name))
self._load_assets(path)
def _load_assets(self, path):
with open('{}/char_indices.txt'.format(path)) as f:
self.char_indices = json.loads(f.read())
self.chars = sorted(self.char_indices.keys())
self.num_chars = len(self.chars)
with open('{}/indices_char.txt'.format(path)) as f:
self.indices_char = json.loads(f.read())
def _sample(self, preds, temperature=.6):
# helper function to sample an index from a probability array
preds = np.asarray(preds).astype('float64')
preds = np.log(preds) / temperature
exp_preds = np.exp(preds)
preds = exp_preds / np.sum(exp_preds)
probas = np.random.multinomial(1, preds, 1)
return np.argmax(probas)
def _predict(self, args_dict):
'''
Generate text based on seed text.
Args:
sentence: Input seed text to kick off generation.
gen_chars: How many characters of text to generate.
Returns generated text
'''
# The model was trained on lowercase text only, and there is no
# provision in the model itself for handling characters that are
# out of vocabulary.
# To compensate, turn everything into lowercase, then check for
# out-of-vocab characters in the result.
sentence, gen_chars = args_dict["sentence"], args_dict["gen_chars"]
sentence = sentence.lower()
for t, char in enumerate(sentence):
if char not in self.char_indices:
print("Bad char {} at position {}".format(char, t))
raise ValueError(
"Unexpected character '{}' at position {}. "
"Only lowercase ASCII characters, spaces, "
"and basic punctuation are supported.".format(char, t))
# The text passed into the model must be exactly SEED_TEXT_LEN
# characters long, or the model will crash. Pad or truncate.
if len(sentence) > SEED_TEXT_LEN:
sentence = sentence[:SEED_TEXT_LEN]
else:
sentence = sentence.rjust(SEED_TEXT_LEN)
generated = ''
with self.graph.as_default():
for i in range(gen_chars):
x = np.zeros((1, SEED_TEXT_LEN, self.num_chars))
for t, char in enumerate(sentence):
x[0, t, self.char_indices[char]] = 1.
preds = self.model.predict(x, verbose=0)[0]
next_index = self._sample(preds)
next_char = self.indices_char[str(next_index)]
generated += next_char
sentence = sentence[1:] + next_char
return generated
