hwang595 / draco Goto Github PK

Hi, I learn that you didn't use GPU in your experiments, so, how long did you take to run one experiment? As I know, using cpu only to train a neural network may take long time.

q = Queue.Queue()
^^^^^^^^^^^
AttributeError: type object 'Queue' has no attribute 'Queue'

Hi, I was recently running your code. When I use 6 workers, and the batch size is set 180, i.e., each workers computes 30 points, the parameters are as following. However, the loss I get is somehow very large even after 5000 iterations (more than 10000). When uncoded scheme used, however, the loss is around 3. Is there any place that I did wrong?

the parameters are as follows:

In Python 3, the Queue module has been renamed to lowercase queue, and the Queue class is part of this module. How we do this without causing much problems. I imported queue then in run_command(argv, quiet=False) function I changed to this q = queue.Queue(). Now am getting a lot of attribute errors.

hi,

only mention of GPU "enable" flag I could see is in Readme, and nowhere in the code

I assume GPU support is not done yet?