Linlp

How to save model for predicting. I do the follow: ` def save_user_model(self, sess, path): builder = tf.compat.v1.saved_model.Builder(path) sig_def = tf.compat.v1.saved_model.predict_signature_def( inputs={'mid_his_batch_ph': self.mid_his_batch_ph, 'mask': self.mask}, outputs={'output': self.user_eb} ) builder.add_meta_graph_and_variables( sess,...

How to speed up for large dataset

3

Hello, my corpus is 700G, is there any way to speed up？

跑bst的时候训练到中途报如下错误，请问作者知道什么问题吗 tensorflow.python.framework.errors_impl.InvalidArgumentError: 4 root error(s) found. (0) Invalid argument: Incompatible shapes: [102,1,1] vs. [1,3,102,102] [[node model/bst/add_2 (defined at data/code/aftonwu/contUndSta/dev/transformer_layer.py:124) ]] [[replica_1/model/bst_1/ExpandDims/_413]] (1) Invalid argument: Incompatible shapes: [102,1,1] vs. [1,3,102,102] [[node...

About the evaluate method

1

It seems there is a bug in the evaluate function? The predict results have been convert to their source tagid, but in the test dataset, target tags are index, so...