something wrong in tutorial_word2vec_basic.py

[liudragonfly]

self.all_params = [embeddings, nce_weights, nce_biases]

The shape of nce_biases is different from the shape of nce_weights.
As a result, tl.files.save_npz(emb_net.all_params...) throws ValueError.

ValueError: could not broadcast input array from shape(vocabulary_size, embedding_size) into shape(vocabulary_size)

[DEKHTIARJonathan]

I confirm this issue, I can reproduce @zsdonghao.

in main_word2vec_basic()
    271             tl.files.save_npz(emb_net.all_params, name=model_file_name + '.npz', sess=sess)

ValueError: could not broadcast input array from shape (50000,128) into shape (50000)

I have opened a new branch to fix the tutorials: https://github.com/tensorlayer/tensorlayer/tree/tutorial_fix

Most of them use old TF API which is not working anymore (tf.flags namely).

Ping @liudragonfly @aMarry

[zsdonghao]

@DEKHTIARJonathan @liudragonfly Hi, I just run the code, it works fine.
I think this error is related to this issue #160

[DEKHTIARJonathan]

@zsdonghao how comes ? I can't manage to make this work...