在运行第一次迭代后,代码会告诉Train_op在使用之前已被引用。我也尝试在if语句之前初始化它,但是出现以下错误
TypeError:train_op必须为Operation或Tensor,给定为:{}
def cnn_model_fn(features, labels, mode):
input_layer = tf.reshape(features["x"], [-1,32,32,3])
conv1 = tf.layers.conv2d(inputs=input_layer, filters=32, kernel_size = [5,5],activation=tf.nn.relu)
pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
conv2 = tf.layers.conv2d(inputs=pool1, filters=64, kernel_size = [5,5],activation=tf.nn.relu)
pool2 = tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
pool2_flat = tf.reshape(pool2, [-1,5 * 5 * 64])
dense = tf.layers.dense(inputs=pool2_flat, units=32, activation=tf.nn.relu)
dropout = tf.layers.dropout(inputs=dense, rate=0.4, training=mode == tf.estimator.ModeKeys.TRAIN)
logits = tf.layers.dense(inputs=dropout, units=2)
#it will provide 2 outputs as cat or dog
predictions = {
"classes": tf.argmax(input=logits, axis=1),
"probabilities": tf.nn.softmax(logits, name="softmax_tensor")
}
if mode == tf.estimator.ModeKeys.PREDICT:
return tf.estimator.EstimatorSpec(mode-mode, predictions=predictions)
#calculatng loss
loss = tf.losses.sparse_softmax_cross_entropy(labels=labels, logits=logits)
#traning op
#train_op= ()
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
train_op = optimizer.minimize(
loss=loss,
global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode=mode, loss=loss, train_op=train_op)
#evaluation matrix
eval_metric_ops={"accuracy" : tf.metrics.accuracy(labels=labels, predictions=predictions["classes"])}
return tf.estimator.EStimatorSpec(mode=mode, loss=loss, eval_metric_ops=eval_metric_ops)
答案 0 :(得分:0)
好像您在train_op
语句中创建if
。是否更改了此内容?
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
train_op = optimizer.minimize(
loss=loss,
global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode=mode, loss=loss, train_op=train_op)
对此:
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
train_op = optimizer.minimize(
loss=loss,
global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode=mode, loss=loss, train_op=train_op)
解决您的问题?
请注意,您的第二个return
是“不可访问的”,即,您将始终在到达第二个return语句之前从该函数返回。