鍦≒ython涓紝tf.placeholder鍑芥暟鐢ㄤ簬瀹氫箟涓€涓彲鍗犱綅鐨則ensor锛屽彲浠ュ湪TensorFlow鐨勮绠楀浘涓綔涓鸿緭鍏ヨ妭鐐广€傚綋鎵ц璁＄畻鍥炬椂锛屽彲浠ラ€氳繃浼犲叆鍏蜂綋鐨勫€兼潵濉厖杩欎釜鍗犱綅绗︺€?/p>

tf.placeholder鍑芥暟鐨勭敤娉曞涓嬶細

tf.placeholder(
    dtype,
    shape=None,
    name=None
)

鍙傛暟璇存槑锛?/p>

• dtype锛氭寚瀹氬崰浣嶇鐨勬暟鎹被鍨嬶紝鍙互鏄痶f.float32銆乼f.float64銆乼f.int32绛夈€?/li>
• shape锛氭寚瀹氬崰浣嶇鐨勫舰鐘讹紝鍙互鏄竴涓垪琛ㄦ垨鍏冪粍锛屼篃鍙互鏄疦one琛ㄧず浠绘剰褰㈢姸銆?/li>
• name锛氭寚瀹氬崰浣嶇鐨勫悕绉帮紝鍙€夊弬鏁般€?/li>

绀轰緥浠ｇ爜锛?/p>

import tensorflow as tf

# 瀹氫箟涓€涓崰浣嶇
x = tf.placeholder(dtype=tf.float32, shape=(None, 2), name='input')

# 鍒涘缓涓€涓绠楀浘
y = tf.reduce_sum(x, axis=1)

# 鍒涘缓涓€涓細璇?/span>
with tf.Session() as sess:
    # 浣跨敤feed_dict鍙傛暟浼犲叆鍏蜂綋鐨勫€兼潵濉厖鍗犱綅绗?/span>
    result = sess.run(y, feed_dict={x: [[1, 2], [3, 4], [5, 6]]})
    print(result)  # 杈撳嚭[3.  7. 11.]

鍦ㄤ笂杩颁唬鐮佷腑锛岄鍏堥€氳繃tf.placeholder瀹氫箟浜嗕竴涓崰浣嶇x锛屾暟鎹被鍨嬩负tf.float32锛屽舰鐘朵负(None, 2)锛岃〃绀哄彲浠ユ帴鍙椾换鎰忚銆?鍒楃殑杈撳叆銆傜劧鍚庯紝鍦ㄨ绠楀浘涓娇鐢ㄤ簡杩欎釜鍗犱綅绗?code>x锛岄€氳繃tf.reduce_sum瀵?code>x鐨勭浜屼釜缁村害杩涜姹傚拰鎿嶄綔銆傛渶鍚庯紝鍦ㄤ細璇濅腑浣跨敤sess.run鎵ц璁＄畻鍥炬椂锛岄€氳繃feed_dict鍙傛暟灏嗗叿浣撶殑鍊?code>[[1, 2], [3, 4], [5, 6]]浼犲叆鍗犱綅绗?code>x锛岃绠楀緱鍒扮粨鏋?code>[3. 7. 11.]銆?/p>