yanwii / dynamic-seq2seq

seq2seq中文聊天机器人

dynamic-seq2seq's Issues

训练报错--assert len(self.data) > 0, ValueError("训练数据为空")

每个回答后面都有__EOS__

每次回答后面都会带一个__EOS__，这是为啥

训练了语料库 但是感觉还是傻傻的

有什么办法让他看起来不是特别傻吗?

你好，请问这个是说我没有训练吗？我看里面有你之前训练的模型，我直接拿来试一试。

Traceback (most recent call last):
File "seq2seq.py", line 160, in
if sys.argv[1]:
IndexError: list index out of range

encoder_vocab_size 如何设置?

encoder_vocab_size设置多少合理?
第一次训练设置7000
但是第二次不设置7000 就会出现问题
这个该如何设置?

是否能提供一個接口餵入pre training的Embedding

您好！
非常感謝您提供這樣如此方便的工具來操作

想問是否能提供一個接口
來放入pre training的word embedding?

你好，运行seq2seq.py，显示出No module named 'chatAction'

训练自己的预料库出现了问题，decoder_label的维度和logits的维度不匹配

只有个别语句出现了这种问题 实在是想不明台为什么

请问你在训练模型的时候训练了多久啊？

我加入一个循环输入之后，运行一直卡在这里，请问大神知道原因吗？

('tensorflow version: ', '1.4.1')
Building prefix dict from the default dictionary ...
Loading model from cache /tmp/jieba.cache
Loading model cost 0.482 seconds.
Prefix dict has been built succesfully.
encoder_vocab_size 1930
decoder_vocab_size 3084
2018-07-24 16:16:14.410099: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1
model/chatbot_seq2seq.ckpt--1

怎么样动态查询天气？

项目中的天气查询怎么可以问一个地方的天气，chatbot自动查询出来呢，求指教大佬

没有过滤掉词频较小的词？

我看preprocess.py 中没有过滤掉词频较小的词，这样数据量一大，词典就很大， 而且很多低频词，这样做seq2seq是不是不太合理？

error

Traceback (most recent call last):
File "seq2seq.py", line 382, in
seq = seq2seq()
File "seq2seq.py", line 54, in init
time_major=True)
File "/Users/LymanYoung/source/git_tensorflow/seq2seq-demo/dynamic-seq2seq/dynamic_seq2seq_model.py", line 61, in init
self._make_graph()
File "/Users/LymanYoung/source/git_tensorflow/seq2seq-demo/dynamic-seq2seq/dynamic_seq2seq_model.py", line 85, in _make_graph
self._init_decoder()
File "/Users/LymanYoung/source/git_tensorflow/seq2seq-demo/dynamic-seq2seq/dynamic_seq2seq_model.py", line 221, in _init_decoder
decoder_fn_train = seq2seq.simple_decoder_fn_train(encoder_state=self.encoder_state)
AttributeError: module 'tensorflow.contrib.seq2seq' has no attribute 'simple_decoder_fn_train'

encoder_state return Tuple type

how to Concatenate the final state? I tried tf.concat((encoder_fw_state, encoder_bw_state), 2), but it return error. thanks.

多轮对话的程序有进展吗

请问，目前项目还维护吗？多轮对话功能有进展吗？

_{Sent with GitHawk}

我使用了小黄鸡的语料，loss值在1左右，但是效果还是很差

迭代5w次左右

想改用自己的训练数据时 出现InvalidArgumentError，不知道是什么原因

假如把语料库改成你的5000句的
即：python prepare_dialog.py 5000
再进行训练 python seq2seq.py retrain
出现这个错误
InvalidArgumentError (see above for traceback): logits and labels must have the same first dimension, got logits shape [2800,5970] and labels shape [2860] [[Node: sequence_loss/SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits = SparseSoftmaxCrossEntropyWithLogits[T=DT_FLOAT, Tlabels=DT_INT32, _device="/job:localhost/replica:0/task:0/device:CPU:0"](sequence_loss/Reshape, sequence_loss/Reshape_1)]]

请问这是什么问题？

生成词向量用的是什么方法啊？

我看着就是词频统计然后给句子做成了向量

[Error] 出現not found model

請問大大我用python3 seq2seq.py train 訓練完成後
輸入 python3 seq2seq.py infer
他都顯示not found model，這是為什麼

模型训练

请问现在项目中的模型是训练好的吗？测试一下效果很差。。。是都需要重新训练模型吗？

训练时，label与logits第一维度不匹配

在进行第一轮训练时，当step101结束后报错：logits and labels must have the same first dimension, got logits shape [2800,9003] and labels shape [2860]，不知道该怎么调整代码的label与logits长度相等。

train error

if my A and Q fiie line over 10, it not good for train, tensorflow.python.framework.errors_impl.InvalidArgumentError: logits and labels must have the same first dimension, got logits shape [20,339] and labels shape [22]

可以提供preprocessing/dec.vocab数据吗？

您好，我对您的这个项目很感兴趣，但是您可以提供preprocessing/dec.vocab, preprocessing/dec.segement， preprocessing/enc.vocab, preprocessing/enc.segement这些数据吗？

非常感谢！

ValueError: Attempt to reuse RNNCell

出现如下错误,请帮忙解答:

ValueError: Attempt to reuse RNNCell <tensorflow.contrib.rnn.python.ops.core_rnn_cell_impl.LSTMCell object at 0x0000000007E92E48> with a different variable scope than its first use. First use of cell was with scope 'BidirectionalEncoder/bidirectional_rnn/fw/lstm_cell', this attempt is with scope 'BidirectionalEncoder/bidirectional_rnn/bw/lstm_cell'. Please create a new instance of the cell if you would like it to use a different set of weights. If before you were using: MultiRNNCell([LSTMCell(...)] * num_layers), change to: MultiRNNCell([LSTMCell(...) for _ in range(num_layers)]). If before you were using the same cell instance as both the forward and reverse cell of a bidirectional RNN, simply create two instances (one for forward, one for reverse). In May 2017, we will start transitioning this cell's behavior to use existing stored weights, if any, when it is called with scope=None (which can lead to silent model degradation, so this error will remain until then.)

大神，请问为什么我训练的数据一直不收敛啊

我按照你的数据去训练，但是没有收敛，