|
@@ -146,7 +146,7 @@ def dynamicRNN(x, seqlen, weights, biases):
|
|
|
# Indexing
|
|
|
outputs = tf.gather(tf.reshape(outputs, [-1, n_hidden]), index)
|
|
|
|
|
|
- # Linear activation, using rnn inner loop last output
|
|
|
+ # Linear activation, using outputs computed above
|
|
|
return tf.matmul(outputs, weights['out']) + biases['out']
|
|
|
|
|
|
pred = dynamicRNN(x, seqlen, weights, biases)
|