Thesis/pytorch-dnc
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

not reliable advise

Browse Source
This commit is contained in:
ixaxaar 2017-11-01 15:08:17 +05:30
parent af1a77ca7f
commit 9e1b0a2983
1 changed files with 0 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
README.md
View File

@ -94,7 +94,6 @@ The visdom dashboard shows memory as a heatmap for batch 0 every `-summarize_fre
## General noteworthy stuff
1. DNCs converge with Adam and RMSProp learning rules, SGD generally causes them to diverge.
2. Using a large batch size (> 100, recommended 1000) prevents gradients from becoming `NaN`.
Repos referred to for creation of this repo: