Embedding configured incorrectly

[ssatia]

self. inputs_with_embed holds the embeddings concatenated with self.inputs. However, the embeddings are not being used since the RRN is still using self.inputs:

stock-rnn/model_rnn.py

Line 102 in aa03e93

val, state_ = tf.nn.dynamic_rnn(cell, self.inputs, dtype=tf.float32, scope="dynamic_rnn")

[lilianweng]

You are totally right. The issue has been fixed in commit 084707d :)

Thanks!