chen feiyu

感觉这个损失函数是不是有点问题啊，tf.reduce_sum(cost) / tf.to_float(batch_size)，分子是cost的和，分母怎么会是batchsize?不应该是tf.reduce_sum(loss_mask)么？

fortest方法

1

https://github.com/bojone/bert4keras/blob/4dcda150b54ded71420c44d25ff282ed30f3ea42/bert4keras/snippets.py#L433 fortest是不是应该以参数的形式从to_dataset中传递进来（和forfit对应）？貌似目前没有用到呢？目前似乎只能逐条测试否则都会打乱？ ` class data_generator_fortest(DataGenerator): """数据生成器 """ def __iter__(self, random=False): for is_end, (text, label) in self.sample(random): token_ids, segment_ids = tokenizer.encode(text, maxlen=maxlen) yield [token_ids, segment_ids], [[label]] # 返回一条样本 def forfit(self, random=False):...

How to train my own data set?

2

Could you please share a pipline for pre-preparing a new data for training?

About Training Time And Gpu

1

Could you please telling me how long did you take in pre-training time，stage1 and stage2 respectively？ And what gpu and how many did you use？I need to pre-train with my...

Why does this white lane line separate the bidirectional traffic in Town07?

1


How to change batch size? I have many images to remove watermarks.

Enforce the traffic lights to operate according to a fixed duration

1

I have known that the red light can be influenced by other traffic light within group, while how to ignore this effect? I meet a scenario that contain only a...