You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello! I'm having a bit of trouble reproducing the results for cyclic HMC with your code. Could you please help me with that?
I'm seeing parser.add_argument('--alpha', type=int, default=1, help='1: SGLD; <1: SGHMC') but shouldn't alpha here be float in [0,1]?
Also, if I allow it to be float and test the training on some value of alpha which seems reasonable (e.g. alpha = 0.05 resulting in momentum term = 0.95 which should be OK?), network weights seem to go to nan soon.
The text was updated successfully, but these errors were encountered:
Hello! I'm having a bit of trouble reproducing the results for cyclic HMC with your code. Could you please help me with that?
I'm seeing
parser.add_argument('--alpha', type=int, default=1, help='1: SGLD; <1: SGHMC')
but shouldn't alpha here be float in [0,1]?Also, if I allow it to be float and test the training on some value of alpha which seems reasonable (e.g. alpha = 0.05 resulting in momentum term = 0.95 which should be OK?), network weights seem to go to nan soon.
The text was updated successfully, but these errors were encountered: