def test_persists_save(model, save_file): model.save_using(tf.train.Saver()) t1 = model.sess.run(model.A) model.save(save_file) m2 = load_tagger_model(save_file) t2 = model.sess.run(m2.A) np.testing.assert_allclose(t1, t2)
def test_persists_save(model, save_file): model.save_using(tf.train.Saver()) t1 = model.sess.run(model.A) model.save(save_file) m2 = load_tagger_model(save_file) t2 = model.sess.run(m2.A) np.testing.assert_allclose(t1, t2)
def _create_model(self, sess, basename, **kwargs): model = load_tagger_model(basename, sess=sess, **kwargs) softmax_output = tf.nn.softmax(model.probs) values, _ = tf.nn.top_k(softmax_output, 1) indices = model.best if self.return_labels: labels = read_json(basename + '.labels') list_of_labels = [''] * len(labels) for label, idval in labels.items(): list_of_labels[idval] = label class_tensor = tf.constant(list_of_labels) table = tf.contrib.lookup.index_to_string_table_from_tensor(class_tensor) classes = table.lookup(tf.to_int64(indices)) return model, classes, values else: return model, indices, values
def _create_model(self, sess, basename, **kwargs): model = load_tagger_model(basename, sess=sess, **kwargs) softmax_output = tf.nn.softmax(model.probs) values, _ = tf.nn.top_k(softmax_output, 1) indices = model.best if self.return_labels: labels = read_json(basename + '.labels') list_of_labels = [''] * len(labels) for label, idval in labels.items(): list_of_labels[idval] = label class_tensor = tf.constant(list_of_labels) table = tf.contrib.lookup.index_to_string_table_from_tensor(class_tensor) classes = table.lookup(tf.to_int64(indices)) return model, classes, values else: return model, indices, values
def load_model(modelname, **kwargs): return load_tagger_model(BASELINE_TAGGER_LOADERS, modelname, **kwargs)
def load_model(modelname, **kwargs): return load_tagger_model(RNNTaggerModel.load, modelname, **kwargs)