Question

当我们声明变量logits时，变量：

fully_connected / weights_1：0，

fully_connected / biases_1：0

已创建。如何在不发生这种情况的情况下恢复日志？

所有全局变量均为：

fully_connected /权重：0， fully_connected /偏压：0， beta1_power：0， beta2_power：0， fully_connected /权重/亚当：0， fully_connected /权重/ Adam_1：0， fully_connected /偏压/亚当：0， fully_connected /偏压/ Adam_1：0， fully_connected / weights_1：0， fully_connected / biases_1：0

ROOT_PATH = "datasets"
directory = TEST_DATA_SET
test_data_dir = os.path.join(ROOT_PATH, directory, "Testing")

# Restore session and variables/nodes/weights
session = tf.Session()
meta_file = os.path.join("output", MODEL_DIR, "save.ckpt.meta")
new_saver = tf.train.import_meta_graph(meta_file)
checkpoint_dir = os.path.join("output", MODEL_DIR)
new_saver.restore(session, tf.train.latest_checkpoint(checkpoint_dir))

# Load the test dataset.
test_images, test_labels = load_data(test_data_dir)

# Transform the images, just like we did with the training set.
test_images32 = [skimage.transform.resize(image, (IMAGE_SCALE_SIZE_X, IMAGE_SCALE_SIZE_Y)) for image in test_images]

# Create a graph to hold the model.
graph = session.graph
#with graph.as_default():
# Placeholders for inputs and labels.
images_ph = tf.placeholder(tf.float32, [None, IMAGE_SCALE_SIZE_X, IMAGE_SCALE_SIZE_Y, 3])

# Flatten input from: [None, height, width, channels]
# To: [None, height * width * channels] == [None, 3072]

images_flat = tf.contrib.layers.flatten(images_ph)

# Fully connected layer.
# Generates logits of size [None, 62]
logits = tf.contrib.layers.fully_connected(images_flat, 62, tf.nn.relu)

predicted_labels = tf.argmax(logits, 1)

Answer 1

想出来。

Logits = tf.nn.relu(tf.matmul(images, weights) + biases)

如何恢复logits

1 个答案: