我正在尝试使用the from_generator interface for the Dataset API注入多个＆＃34;轮次＆＃34;输入到图表中。
在first attempt上，我使用repeat() function使发生器连续运行3次。但是，batch_join call批量大小不是每轮迭代次数的偶数倍（批量大小为3次的10次迭代），来自不同＆＃34;的数据。轮＆＃34; /＆＃34; epochs＆＃34;最终在同一批次中（取决于张量的处理顺序;图中有一些并行性）。
在我的second attempt上，我尝试在每个纪元完成后重新运行迭代器。但是，只要tf.errors.OutOfRangeError is thrown sess.run() on the output of the batch call，rerunning the iterator's initializer的所有后续调用都会再次抛出OutOfRangeError，即使在shortcut methods to embed YouTube video (works on mobile too) See here之后。
我想将多轮输入连续注入图形中，而不是像第一个例子那样重叠它们（例如在批处理选项上使用allow_smaller_final_batch）。我在自定义Tensorflow fork中实例化的一些内核重启非常昂贵，例如mmap一个O（10gb）的文件，所以我想以某种方式充分利用这两个世界。
答案 0 :(得分：2)
I think the problem stems from using template <class T>
T* LoadableObject::instantiate(const CreationData& cd)
{
T* obj = new(T);
if (!obj->initialize(cd))
{
delete(obj);
obj = NULL;
}
return obj;
}
(which supports reinitialization) with <a :href="'https://link-to-whatever.com/'+ (diced==6 ? 'winner' : 'looser')">LINK</a>
(which uses TensorFlow queues and queue-runners, and hence does not support reinitialization).
I'm not completely clear what your code is doing, but I think you can implement the entire pipeline as a struct Employee
{
public string FirstName { get; set; }
public string LastName { get; set; }
public string FullName
{
get => FirstName + LastName;
}
public int Age { get; set; }
public string Role { get; set; }
public double Salary { get; set; }
public bool IsManager { get; set; }
public static void EditInfo(ref Employee e1, ref Employee e2)
{
if (e1.IsManager)
{
Console.WriteLine($"Feel free to set {e2.FirstName}'s info: ");
e2.FirstName = Console.ReadLine();
e2.LastName = Console.ReadLine();
Console.WriteLine($"From now on thy name is {e2.FullName}!");
}
else if (e2.IsManager)
{
Console.WriteLine($"Feel free to set {e1.FirstName}'s info: ");
e1.FirstName = Console.ReadLine();
e1.LastName = Console.ReadLine();
Console.WriteLine($"From now on thy name is {e1.FullName}!");
}
}
}
. Replace the following fragment of code:
tf.contrib.data.Dataset...with something like the following:
tf.train.batch_join()