answers.append(answer)
print(answer)
return np.array(alignments), test_source_text.split(' '), out_words
if not os.path.exists('checkpoints_luong/encoder'):
os.makedirs('checkpoints_luong/encoder')
if not os.path.exists('checkpoints_luong/decoder'):
os.makedirs('checkpoints_luong/decoder')
# Uncomment these lines for inference mode
encoder_checkpoint = tf.train.latest_checkpoint('checkpoints_luong/encoder')
decoder_checkpoint = tf.train.latest_checkpoint('checkpoints_luong/decoder')
if encoder_checkpoint is not None and decoder_checkpoint is not None:
encoder.load_weights(encoder_checkpoint)
decoder.load_weights(decoder_checkpoint)
if MODE == 'train':
for e in range(NUM_EPOCHS):
en_initial_states = encoder.init_states(BATCH_SIZE)
encoder.save_weights(
'checkpoints_luong/encoder/encoder_{}.h5'.format(e + 1))
decoder.save_weights(
'checkpoints_luong/decoder/decoder_{}.h5'.format(e + 1))
for batch, (source_seq, target_seq_in,
target_seq_out) in enumerate(dataset.take(-1)):
loss = train_step(source_seq, target_seq_in, target_seq_out,
en_initial_states)
if batch % 100 == 0:
out_words = []
alignments = []
while True:
de_output, de_state_h, de_state_c, alignment = decoder(
de_input, (de_state_h, de_state_c), en_outputs[0])
de_input = tf.expand_dims(tf.argmax(de_output, -1), 0)
out_words.append(fr_tokenizer.index_word[de_input.numpy()[0][0]])
alignments.append(alignment.numpy())
if out_words[-1] == '<end>' or len(out_words) >= 20:
break
dist = ' '.join(out_words)
dists.append(dist)
return np.array(alignments), test_source_text.split(' '), out_words
encoder.load_weights('checkpoints_luong/encoder/encoder.h5')
decoder.load_weights('checkpoints_luong/decoder/decoder.h5')
for i, test_sent in enumerate(test_sents):
test_sequence = normalize_string(test_sent)
alignments, source, prediction = predict(test_sequence)
if i % 100 == 0: print(i, ' Predictions completed!')
dists = ["'" + d[:-6] + "'" for d in dists]
r['distractor'] = dists
r.to_csv('Predictions.csv', index=None)
try:
with open(file, 'r') as csvfile:
reader = csv.reader(csvfile)
table = [[float(e) for e in r] for r in reader]
table = tf.cast(numpy.asarray(table), dtype=tf.dtypes.float32)
return table
except Exception:
return None
embedding = read_table("random_embedding.csv")
dialog_list = Dialog.resolve_data("./data/movie_conversations.txt",
"./data/movie_lines.txt",
vocab_size=800)
Dialog.load_word2ids()
encoder = Encoder(batch_size=1, units=64, embedding=embedding)
decoder = Decoder(batch_size=3, units=64, embedding=embedding, vocab_size=804)
encoder.load_weights('encoder_weights_saving')
decoder.load_weights('decoder_weights_saving')
decoder.training = False
decoder.mode = "beam"
print("<<< Chatting Bot Demonstration >>>")
print("Mode: 3 - Beam Search")
val = ""
while val != "Exit":
val = input("Me: ")
inp = str("TSTSTARTTST " + val + " TSTEOSTST")
rep = predict(inp)
print("Computer: " + rep)
