我尝试使用此功能将日志文本写入Tensorboard日志文件,但我遇到了一些麻烦。
我想写一个包含字符串的列表(或np.array)。我不能通过这个,因为它不是一个张量,所以我该怎么办呢?
我正在尝试:
hyperparameters = ["learning_rate=1","batch_size=50","optimizer=Adagrad"]
summary_op = tf.summary.text("hyperparameters info", hyperparameters)
summary = session.run(summary_op )
writer.add_summary(summary)
请注意我已经定义了FileWriter并且我已经在运行会话。
答案 0 :(得分:1)
你可以使用这样的张量:
hyperparams = np.array(["learning_rate=1","batch_size=50","optimizer=Adagrad"])
hyperparams_tensor = tf.constant(hyperparams)
#Or Directly use the tensor, there is no need for np.array() or list
#hyperparams_tensor = tf.constant(["learning_rate=1","batch_size=50","optimizer=Adagrad"])
summary_op = tf.summary.text("hyperparameters info", hyperparams_tensor)
summary = session.run(summary_op)
writer.add_summary(summary)