search

keithito / tacotron

A TensorFlow implementation of Google's Tacotron speech synthesis with pre-trained model (unofficial)
MIT License
2.94k stars 965 forks source link

ask for help, load model by tensor-name and sess.run , no response #279

Closed Liujingxiu23 closed 5 years ago

Liujingxiu23 commented 5 years ago

load model as follows, no error, no response, I can not finger out anything wrong. Ask for help I try this since I want to change the saved model(check-point) to another format using tf.saved_model.builder.SavedModelBuilder.

with tf.Session(config=config) as sess: saver = tf.train.import_meta_graph(meta_path) saver.restore(sess, checkpoint_path) x = tf.get_default_graph().get_tensor_by_name("datafeeder/inputs:0") x_len = tf.get_default_graph().get_tensor_by_name("datafeeder/input_lengths:0") y = tf.get_default_graph().get_tensor_by_name("model/inference/dense_1/BiasAdd:0") y_out = sess.run(y, feed_dict = {x:inputs, x_len:inputs_length}

Liujingxiu23 commented 5 years ago

Solved

xxoospring commented 5 years ago

i also met this problem, could you tell me the reason?