jianli-Alex

**bug描述**  `state = None`放在了`for epoch in range(num_epochs)`上. 因此,在新的epoch中,state实际上没有被初始化,而是使用上一epoch最后一次迭代最后一个时间步的state, 这样并不合理. 要把`state = None`放在了`for epoch in range(num_epochs)`下面,`for X, Y in data_iter`上面 **版本信息** pytorch: 1.6.0 torchvision: 0.7.0 torchtext: ...

**bug描述** ``` # 本函数已保存在d2lzh_pytorch包中方便以后使用 def data_iter_random(corpus_indices, batch_size, num_steps, device=None): # 减1是因为输出的索引x是相应输入的索引y加1 num_examples = (len(corpus_indices) - 1) // num_steps epoch_size = num_examples // batch_size example_indices = list(range(num_examples)) random.shuffle(example_indices) # 返回从pos开始的长为num_steps的序列 def...

Some Question about "Mean" image vector

10

Hello, I found that the image data have been extracted by VGG and change in .npy file. But I have a question in "MEAN" image in vistaNet. How do you...