Text-Generate-RNN icon indicating copy to clipboard operation
Text-Generate-RNN copied to clipboard

Published 20 hours ago • verified publisher icon stardut

能提供下模型文件嘛

Open xxllp opened this issue 5 years ago • 21 comments

自己训练感觉要好久的

xxllp avatar Dec 17 '18 02:12 xxllp

邮箱，晚些时候发给你

stardut avatar Dec 17 '18 06:12 stardut

我训练到loss=2左右 看predict 还是很一般，不知道最后的效果是否能有多好

xxllp avatar Dec 17 '18 06:12 xxllp

InvalidArgumentError (see above for traceback): Assign requires shapes of both tensors to match. lhs shape= [7773,512] rhs shape= [7773,128] [[Node: save/Assign_3 = Assign[T=DT_FLOAT, _class=["loc:@rnn/embedding"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](rnn/embedding/Adam, save/RestoreV2_3/_19)]]

xxllp avatar Dec 17 '18 08:12 xxllp

我拿自己的模型 predict 一直报错。。

xxllp avatar Dec 17 '18 08:12 xxllp

不好意思，最近很忙，现在才回。你将sample.py和train.py的num_units 数量调到一致，应该就ok了，如果你已经训练了模型，那就将sample.py的num_units 调到和train.py一样。

stardut avatar Dec 21 '18 03:12 stardut

模型还是发一份给我吧，

xxllp avatar Dec 21 '18 03:12 xxllp

你的邮箱？

stardut avatar Dec 21 '18 03:12 stardut

[email protected] ，谢谢哈~自己训练半天看不到一个合理的结果

xxllp avatar Dec 21 '18 03:12 xxllp

@xxllp , how long did you take to trained the model ? it took me more than 25 hours and still running . I quit it . what is the accuracy are you getting?

niranjan8129 avatar Dec 24 '18 01:12 niranjan8129

我运行了几个小时看loss 没有下降就停止了。大概1-2左右吧

xxllp avatar Dec 24 '18 01:12 xxllp

@xxllp you meant to say that you trained the model successfully for few hours . what's is going on on my machine? model is still running .. almost over 26 hours ..no result

you said " loss stop without falling" , you got loss 0.0 after running 1-2 hours and stopped without falling

niranjan8129 avatar Dec 24 '18 01:12 niranjan8129

@xxllp did you modify the parameters values to be completed/stop automatically? if so, could you please help me what the parameter values you modified

niranjan8129 avatar Jan 04 '19 05:01 niranjan8129

作者不开源模型，我也没搞了

xxllp avatar Jan 04 '19 05:01 xxllp

这个DSSM没有 embedding啊，直接就接了一个全连接层，有问题吧？

ustcJin avatar Jan 24 '19 06:01 ustcJin

可以把模型发我邮箱吗 训练了好久都没停，想看看结果 [email protected]

nsiakjdw avatar Feb 26 '19 09:02 nsiakjdw

@nsiakjdw 模型会自动保存，你可以主动停下来的，我之前训练的模型现在找不到了，那还是我在学校的时候训练的，过去太久没找到（-_-||

stardut avatar Feb 26 '19 09:02 stardut

哈哈 解决了 就是模型加载那里有点小问题，谢谢了

nsiakjdw avatar Feb 26 '19 13:02 nsiakjdw

@nsiakjdw @stardut thank you. could you please upload latest code ?

niranjan8129 avatar Feb 26 '19 13:02 niranjan8129

@nsiakjdw @stardut thank you. if you do not have issue, could you please share code with me?

niranjan8129 avatar Mar 01 '19 05:03 niranjan8129

could you please send me the model file? thanks! [email protected]

danielkaifeng avatar Mar 01 '19 11:03 danielkaifeng

I used stardust code(stardut/Text-Generate-RNN) but could not trained it and did not stop it . could you please provide me the latest code if you have ?

niranjan8129 avatar Mar 01 '19 13:03 niranjan8129