Chujie Zeng

For Python3, you should code like this: ``` if FLAGS.word2vec: # initial matrix with random uniform initW = np.random.uniform(-0.25,0.25,(len(vocab_processor.vocabulary_), FLAGS.embedding_dim)) # load any vectors from the word2vec print("Load word2vec file...

Here is what I did. ``` # Final (unnormalized) scores and predictions with tf.name_scope("output"): self.output_W = tf.get_variable( "W", shape=[num_filters_total, num_classes], initializer=tf.contrib.layers.xavier_initializer()) b = tf.Variable(tf.constant(0.1, shape=[num_classes]), name="b") l2_loss += tf.nn.l2_loss(self.output_W) l2_loss...

Those codes for clipping W is in file text_cnn.py. So I think clipping is performed every step, at least it's what I want. I might be wrong cos I'm not...

@hkhatod This is called L2-Normalization. The general idea is to keep structural risk minimal.

@csyanbin Agree. In some papers, they claim clipping is not elegant. It's more like an empirical trick.

On my MacBook Air with only 4 GM RAM, it takes me 2.5 hours. You might increase batch size to a much greater number instead of 64 since you have...

@phpmind Holy shh... That's really a MONSTER... 👍

Hi @DoctorWkt . Probably just mentioning the changes in readme would be great.

你的是啥...我的1060不知道能不能刚动 Deep Learning课的final proj要刚一下一个差不多的数据集....

OMG So it will take me whole life to finish my final proj. I should turn to fine-tuning I think...