OrgLion-ML
/
dlib
mirror of https://github.com/davisking/dlib.git
1
0
Fork 0
Code Issues Releases Wiki Activity

Clarified spec

This commit is contained in:
Davis King 2015-12-08 08:25:29 -05:00
parent 5f8e41a889
commit 3222a3af6b
1 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
dlib/dnn/layers_abstract.h
View File

@ -194,8 +194,8 @@ namespace dlib
- setup() has been called. - setup() has been called.
- computed_output is the tensor resulting from calling forward(sub,computed_output). - computed_output is the tensor resulting from calling forward(sub,computed_output).
Moreover, this was the most recent call to forward(). This means that Moreover, this was the most recent call to forward(). This means that
backward() is allowed to cache intermediate results computed during forward() is allowed to cache intermediate results so they can be used
forward() and use them for the backward computation. during the backward computation.
- have_same_dimensions(gradient_input, computed_output) - have_same_dimensions(gradient_input, computed_output)
- have_same_dimensions(sub.get_gradient_input(), sub.get_output()) == true - have_same_dimensions(sub.get_gradient_input(), sub.get_output()) == true
- have_same_dimensions(params_grad, get_layer_params()) == true - have_same_dimensions(params_grad, get_layer_params()) == true
@ -248,9 +248,8 @@ namespace dlib
requires requires
- setup() has been called. - setup() has been called.
- computed_output is the tensor resulting from the most recent call to - computed_output is the tensor resulting from the most recent call to
forward_inplace(). This means that backward_inplace() is allowed to forward_inplace(). This means that forward_inplace() is allowed to cache
cache intermediate results computed during forward_inplace() and use them intermediate results so they can be used during the backward computation.
for the backward computation.
- have_same_dimensions(gradient_input, data_grad) == true - have_same_dimensions(gradient_input, data_grad) == true
- have_same_dimensions(gradient_input, computed_output) == true - have_same_dimensions(gradient_input, computed_output) == true
- have_same_dimensions(params_grad, get_layer_params()) == true - have_same_dimensions(params_grad, get_layer_params()) == true