Welcome to Constant Etherpad!

These pads are archived each night (around 4AM CET) @ http://etherdump.constantvzw.org/
An RSS feed from the etherdump also appears on http://constantvzw.org/

To prevent your public pad from appearing in the archive and RSS feed, put or just leave the following (including the surrounding double underscores) anywhere in the text of your pad:

    __NOPUBLISH__

Changes will be reflected after the next update at 4AM.
installeren van Torch-RNN
http://pad.constantvzw.org/p/algorithmic_uncertainty.torch-rnn.install
code: https://github.com/jcjohnson/torch-rnn

run Torch op de computer van Gijs
Omdat ik inlog over ssh op de server, maar het proces wil laten draaien gebruik ik 'screen'

ssh algolit@UVW

screen -r

Verzeker dat de venv geactiveerd is
. ~/venvs/torch-rnn/bin/activate

(her-)starten training:
th train.lua -input_h5 data/tiny-shakespeare.h5 -input_json data/tiny-shakespeare.json -model_type rnn -num_layers 2 -rnn_size 256 -wordvec_size 128


Sample nemen van getrained model:
th sample.lua -checkpoint cv/checkpoint_89000.t7 -length 200

Keuze voor Torch-RNN
Oorspronkelijk idee: http://karpathy.github.io/2015/10/25/selfie/
Nieuwe, snellere implementatie: https://github.com/karpathy/char-rnn
Motivatie voor de keuze voor torch-rnn

• "Update: torch-rnn
• Justin Johnson  (@jcjohnson) recently re-implemented char-rnn from scratch with a much  nicer/smaller/cleaner/faster Torch code base. It's under the name torch-rnn.  It uses Adam for optimization and hard-codes the RNN/LSTM  forward/backward passes for space/time efficiency. This also avoids  headaches with cloning models in this repo. In other words, torch-rnn  should be the default char-rnn implemention to use now instead of the  one in this code base."

> maar geschreven in LUA

vervolg
Dit proces is vervolgd op http://pad.constantvzw.org/p/algorithmic_uncertainty.code-oefening