Changing a fixed dropout rate of a stored Tensorflow model | بلاگ

Changing a fixed dropout rate of a stored Tensorflow model

تعرفه تبلیغات در سایت

آخرین مطالب

امکانات وب

Vote count: 0

I have a stored Tensorflow model, which has been trained for a long period. For dropout ops, I used a fixed value for the "keep_prob" parameter instead of a placeholder, which now turns out as a mistake, since for evaluation I would like to set the value of "keep_prob" for 1.0.

Is there a way to change this op, to operate with keep_prob = 1.0, for the stored model? In other words, is it possible to change a fixed dropout rate of a stored Tensorflow model?

Specifically, I used the fixed value in these ops:

  • tf.nn.dropout
  • tf.nn.bidirectional_dynamic_rnn
  • tf.nn.dynamic_rnn
asked 41 secs ago
Mega

...
نویسنده : استخدام کار بازدید : 14 تاريخ : چهارشنبه 25 بهمن 1396 ساعت: 1:25