roomo7time

This should be closed as both pooling are the same now.

me thirteen at tensorflow-gpu 1.11.0 for `tf.assert_variables_initialized([V,g,b])`

I am not sure this is a proper resolution for the above, but I tried the below: (a) change `g = tf.get_variable('g', dtype=tf.float32, initializer=scale_init, trainable=True)` by `g = tf.get_variable('g', shape=scale_init.get_shape().as_list(),...

For my cace, I was experiencing the cpu-gpu difference issue with faiss-gpu==1.7.2 that I installed with pip. After uninstalling it, and reintalling faiss-gpu by conda using `conda install -c pytorch...

I got 66% accuracy.