Tensorflow2.0目前是否全面支持？

[hezhefly]

如题，我目前的运行环境使用是tf2.0版本，keras是为2.3.0。 但无法执行 报错信息如下：

~/anaconda3/lib/python3.7/site-packages/keras/engine/training.py in _prepare_total_loss(self, masks)
    690 
    691                     output_loss = loss_fn(
--> 692                         y_true, y_pred, sample_weight=sample_weight)
    693 
    694                 if len(self.outputs) > 1:

TypeError: __call__() got an unexpected keyword argument 'sample_weight'

[uduse]

matchzoo losses do not support sample_weight.

[hezhefly]

matchzoo losses do not support sample_weight.

是啊，那目前为止是不支持keras2.3.0了，是吧？

[uduse]

这和 keras2.3.0 有什么关系吗？

[hezhefly]

这和 keras2.3.0 有什么关系吗？

keras升级到2.3.0会报错

[abelx]

same error tensorflow==1.14.0 keras==2.3.0

[uduse]

@hezhefly @abelx Please provide a short snippet of code to reproduce the error, so I can investigate.

[bwanglzu]

@hezhefly May I know which loss function u're using? How can we reproduce the error?

[matthew-z]

Compile a ranking model using keras 2.3, and you will get this error.

import matchzoo as mz
model = mz.models.DSSM()
model.params["task"] = mz.tasks.Ranking(mz.losses.RankHingeLoss())
model.guess_and_fill_missing_params()
model.build()
model.compile()

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-7-6bbe396c20b1> in <module>
      3 model.guess_and_fill_missing_params()
      4 model.build()
----> 5 model.compile()

~/.anaconda3/envs/python37/lib/python3.7/site-packages/matchzoo/engine/base_model.py in compile(self)
    218         """
    219         self._backend.compile(optimizer=self._params['optimizer'],
--> 220                               loss=self._params['task'].loss)
    221 
    222     def fit(

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py in symbolic_fn_wrapper(*args, **kwargs)
     73         if _SYMBOLIC_SCOPE.value:
     74             with get_graph().as_default():
---> 75                 return func(*args, **kwargs)
     76         else:
     77             return func(*args, **kwargs)

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/engine/training.py in compile(self, optimizer, loss, metrics, loss_weights, sample_weight_mode, weighted_metrics, target_tensors, **kwargs)
    227         #                   loss_weight_2 * output_2_loss_fn(...) +
    228         #                   layer losses.
--> 229         self.total_loss = self._prepare_total_loss(masks)
    230 
    231         # Functions for train, test and predict will

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/engine/training.py in _prepare_total_loss(self, masks)
    690 
    691                     output_loss = loss_fn(
--> 692                         y_true, y_pred, sample_weight=sample_weight)
    693 
    694                 if len(self.outputs) > 1:

TypeError: __call__() got an unexpected keyword argument 'sample_weight'

keras 2.3 forced all the losses to have sample_weight param: check https://github.com/keras-team/keras/blob/1cf5218edb23e575a827ca4d849f1d52d21b4bb0/keras/engine/training.py#L692

or this commit

https://github.com/keras-team/keras/commit/910e1247b452f29c0297ec6e3bf52b28b24fefcf

or API changes:

https://github.com/keras-team/keras/releases

Also, from 2.3, keras.backend.tf also becomes invalid. If you replace the DSSM in the above with KNRM, you will get this error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-9-1315debcec8a> in <module>
      3 model.params["task"] = mz.tasks.Ranking(mz.losses.RankHingeLoss())
      4 model.guess_and_fill_missing_params()
----> 5 model.build()
      6 model.compile()

~/.anaconda3/envs/python37/lib/python3.7/site-packages/matchzoo/models/knrm.py in build(self)
     68                 sigma = self._params['exact_sigma']
     69                 mu = 1.0
---> 70             mm_exp = self._kernel_layer(mu, sigma)(mm)
     71             mm_doc_sum = keras.layers.Lambda(
     72                 lambda x: K.tf.reduce_sum(x, 2))(mm_exp)

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py in symbolic_fn_wrapper(*args, **kwargs)
     73         if _SYMBOLIC_SCOPE.value:
     74             with get_graph().as_default():
---> 75                 return func(*args, **kwargs)
     76         else:
     77             return func(*args, **kwargs)

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/engine/base_layer.py in __call__(self, inputs, **kwargs)
    487             # Actually call the layer,
    488             # collecting output(s), mask(s), and shape(s).
--> 489             output = self.call(inputs, **kwargs)
    490             output_mask = self.compute_mask(inputs, previous_mask)
    491 

~/.anaconda3/envs/python37/lib/python3.7/site-packages/keras/layers/core.py in call(self, inputs)
    298 
    299     def call(self, inputs):
--> 300         return self.activation(inputs)
    301 
    302     def get_config(self):

~/.anaconda3/envs/python37/lib/python3.7/site-packages/matchzoo/models/knrm.py in kernel(x)
     91 
     92         def kernel(x):
---> 93             return K.tf.exp(-0.5 * (x - mu) * (x - mu) / sigma / sigma)
     94 
     95         return keras.layers.Activation(kernel)

AttributeError: module 'keras.backend' has no attribute 'tf'

[aszhanghuali]

@matthew-z The same question! AttributeError: module 'keras.backend' has no attribute 'tf' Have you solved it? Thank you, I look forward to your reply.

[matthew-z]

Please upgrade MatchZoo. The latest version has supported Keras 2.3

On Thu, Oct 24, 2019 at 8:12 PM as15538191859 notifications@github.com wrote:

@matthew-z https://github.com/matthew-z The same question! AttributeError: module 'keras.backend' has no attribute 'tf' Have you solved it? Thank you, I look forward to your reply.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NTMC-Community/MatchZoo/issues/789?email_source=notifications&email_token=ABLZV5ZG2V3QS5BDJII43H3QQGGLTA5CNFSM4IZHGIQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOECEZUOA#issuecomment-545888824, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABLZV56E5YYOVXZ4OGYGIVDQQGGLTANCNFSM4IZHGIQQ .

[aszhanghuali]

First of all, thank you for your reply! Secondly, my MatchZoo's version is 2.1.0

------------------ 原始邮件 ------------------ 发件人: "Z ZH"notifications@github.com; 发送时间: 2019年10月24日(星期四) 晚上8:14 收件人: "NTMC-Community/MatchZoo"MatchZoo@noreply.github.com; 抄送: "红楼小梦"1319466165@qq.com;"Comment"comment@noreply.github.com; 主题: Re: [NTMC-Community/MatchZoo] Tensorflow2.0目前是否全面支持？ (#789)

Please upgrade MatchZoo. The latest version has supported Keras 2.3

On Thu, Oct 24, 2019 at 8:12 PM as15538191859 notifications@github.com wrote:

@matthew-z https://github.com/matthew-z The same question! AttributeError: module 'keras.backend' has no attribute 'tf' Have you solved it? Thank you, I look forward to your reply.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NTMC-Community/MatchZoo/issues/789?email_source=notifications&email_token=ABLZV5ZG2V3QS5BDJII43H3QQGGLTA5CNFSM4IZHGIQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOECEZUOA#issuecomment-545888824, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABLZV56E5YYOVXZ4OGYGIVDQQGGLTANCNFSM4IZHGIQQ .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[matthew-z]

current version is 2.2.0

[hezhefly]

都放弃了，我把matchzoo从我的工程项目中单独抽离出来，打包在docker里运行了

[aszhanghuali]

Thank you for your reply. Why can't I upgrade to 2.2.0, and the latest version is 2.1.0. Can you tell me how you operate? Thank you again！

------------------ 原始邮件 ------------------ 发件人: "Z ZH"notifications@github.com; 发送时间: 2019年10月24日(星期四) 晚上8:25 收件人: "NTMC-Community/MatchZoo"MatchZoo@noreply.github.com; 抄送: "你一定会成为你想的那个人"1319466165@qq.com;"Comment"comment@noreply.github.com; 主题: Re: [NTMC-Community/MatchZoo] Tensorflow2.0目前是否全面支持？ (#789)

current version is 2.2.0

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[bwanglzu]

Guess we forgot to upload to pypi after 2.2 realse..

[bwanglzu]

Sorry for the inconvenience.. we'll do it now.

[aszhanghuali]

font{
    line-height: 1.6;
}
ul,ol{
    padding-left: 20px;
    list-style-position: inside;
}

    Thank you for doing it! I wanna to know what you are going to do.

                            1319466165

                                1319466165@qq.com

    签名由
    网易邮箱大师
    定制

在2019年10月24日 21:00，Wang Bo<notifications@github.com> 写道： 

Sorry for the inconvenience.. I'll do it now.

—You are receiving this because you commented.Reply to this email directly, view it on GitHub, or unsubscribe.

[bwanglzu]

matchzoo 2.2 is on pypi, please:

pip install matchzoo --upgrade