This folder and this README.md
contain some very useful scripts.
The folder content is the following:
check_exp_diff.sh*
: shows difference between experiments' training CLI arguments;update_experiments.sh*
: synchronises experiments across multiple nodes;show_error.plt*
: show current error with gnuplot byawk
-inglast/train.log
;show_error_exp.plt*
: show losses for a specific experiment;image_plot.py
: saves intermediate training visual outputs to PDF (may be used bymain.py
);
To compare the CLI arguments used in two different experiments run ./check_exp_diff.sh <base exp> <new exp>
.
The script will output the --word-diff
between the calling arguments, and git show
the newer experiment's HEAD
.
Replace <{base,new} exp>
with the corresponding three-digit identifier.
To plot the (MSE, CE, replica MSE, and periodic CE) cost functions interactively for the current training network run ./show_error.plt -i
.
To statically plot them, after training, run ./show_error.plt
.
If the experiment data is not on the current machine run ./update_experiments.sh -i
to iteratively sync the data every 5
seconds.
To view the losses iteratively run ./show_error_exp.plt <exp> -i
, where <exp>
is for example 012
.
To run statically restrain to use the -i
flag.
To sync experiments across multiple nodes run ./update_experiments.sh
.
If you whish to run it quietly (verbose is default), run ./update_experiments.sh -q
.
If you'd like to run it in a loop of 5
seconds, run ./update_experiments.sh -i
.
convert *.pdf *.png
Create animations from the content of 00{6,7}/PDFs
into anim/
:
for p in 006/PDFs/*; do
g=anim/${p##*/}
convert -delay 100 $p ${p/6/7} ${g/pdf/gif}
done
delay
is expressed in 10
ms.