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
Copy file name to clipboardExpand all lines: README.md
+8Lines changed: 8 additions & 0 deletions
Original file line number
Diff line number
Diff line change
@@ -62,6 +62,14 @@ optnet = require 'optnet'
62
62
optnet.optimizeMemory(net, input, opts)
63
63
```
64
64
65
+
### Optional parameters
66
+
67
+
Here is a list of options that are currently supported, and should be passed in the `opts` table as a third argument:
68
+
*`inplace`: uses in place modules when available (boolean)
69
+
*`mode`: selects between `training` and `evaluation` optimization algorithm (string)
70
+
*`reuseBuffers`: shares internal buffers between same modules (like unfolded images for convolution) (boolean)
71
+
*`removeGradParams`: remove `gradWeight` and `gradBias` in the networks, saving their sharings so that they can be exactly reconstructed. Only applies for `evaluation` mode (boolean)
72
+
65
73
## Visualizing the memory reuse
66
74
67
75
We can analyse the sharing of the internal buffers by looking at the computation
0 commit comments