Can anyone share the trained model, thanks！

[wsyfh15]

My computer runs this model too slowly. Can anyone share the trained model, thanks！

[wsyfh15]

我把epoch设为3 跑了半天，然后测试的时候什么都检测不出来………………

[ghost]

@wsyfh15 @calmisential 我训练时，第一个epoch loss非常大，到第二个epoch loss一开始变为3，后来又突变到和epoch＝1时好几万的loss了。不知道是什么原因

[wangzyon]

My computer runs this model too slowly. Can anyone share the trained model, thanks！

cython to speed up,such as iou

[zcdliuwei]

@wsyfh15 @calmisential @ghost 我训练时也是损失值忽然增加到几万甚至几十万，请问你们解决这个问题了吗

[zcdliuwei]

@wsyfh15 @calmisential 我训练时，第一个epoch loss非常大，到第二个epoch loss一开始变为3，后来又突变到和epoch＝1时好几万的loss了。不知道是什么原因

我的情况和你一样，亏损忽大忽小，大到几十万，小到个位数，这是什么原因，请问你解决了吗

[zcdliuwei]

作者可以解决一下训练时loss值非常非常大，而且不稳定的问题吗

[lulmer]

Hello, I really liked your code which is very clear but I tried to train your model with collab pro during 12hours and I found it very long since I only reached epoch 7/50. How to speed it up? I tried to put a @tf.function decorator upon your train_step function but there's a lot of change to make in the code in order to make it work. Any pieces of advice? Thank you

[calmiLovesAI]

Hello, I really liked your code which is very clear but I tried to train your model with collab pro during 12hours and I found it very long since I only reached epoch 7/50. How to speed it up? I tried to put a @tf.function decorator upon your train_step function but there's a lot of change to make in the code in order to make it work. Any pieces of advice? Thank you

I will rewrite the code in the last one or two weeks.