def testPostProcessImageTrainMode(self, likelihood, num_mixtures, depth): batch = 1 rows = 8 cols = 24 hparams = hparam.HParams( hidden_size=2, likelihood=likelihood, mode=tf.estimator.ModeKeys.TRAIN, num_mixtures=num_mixtures, ) inputs = tf.random_uniform([batch, rows, cols, hparams.hidden_size], minval=-1., maxval=1.) outputs = common_image_attention.postprocess_image( inputs, rows, cols, hparams) self.assertEqual(outputs.shape, (batch, rows, cols, depth))
def testPostProcessImageTrainMode(self, likelihood, num_mixtures, depth): batch = 1 rows = 8 cols = 24 hparams = tf.contrib.training.HParams( hidden_size=2, likelihood=likelihood, mode=tf.estimator.ModeKeys.TRAIN, num_mixtures=num_mixtures, ) inputs = tf.random_uniform([batch, rows, cols, hparams.hidden_size], minval=-1., maxval=1.) outputs = common_image_attention.postprocess_image( inputs, rows, cols, hparams) self.assertEqual(outputs.shape, (batch, rows, cols, depth))
def testPostProcessImageInferMode(self, likelihood, num_mixtures, depth): batch = 1 rows = 8 cols = 24 block_length = 4 block_width = 2 hparams = hparam.HParams( block_raster_scan=True, hidden_size=2, likelihood=likelihood, mode=tf.estimator.ModeKeys.PREDICT, num_mixtures=num_mixtures, query_shape=[block_length, block_width], ) inputs = tf.random_uniform([batch, rows, cols, hparams.hidden_size], minval=-1., maxval=1.) outputs = common_image_attention.postprocess_image( inputs, rows, cols, hparams) num_blocks_rows = rows // block_length num_blocks_cols = cols // block_width self.assertEqual(outputs.shape, (batch, num_blocks_rows, num_blocks_cols, block_length, block_width, depth))
def testPostProcessImageInferMode(self, likelihood, num_mixtures, depth): batch = 1 rows = 8 cols = 24 block_length = 4 block_width = 2 hparams = tf.contrib.training.HParams( block_raster_scan=True, hidden_size=2, likelihood=likelihood, mode=tf.contrib.learn.ModeKeys.INFER, num_mixtures=num_mixtures, query_shape=[block_length, block_width], ) inputs = tf.random_uniform([batch, rows, cols, hparams.hidden_size], minval=-1., maxval=1.) outputs = common_image_attention.postprocess_image( inputs, rows, cols, hparams) num_blocks_rows = rows // block_length num_blocks_cols = cols // block_width self.assertEqual(outputs.shape, (batch, num_blocks_rows, num_blocks_cols, block_length, block_width, depth))