Skip to content

jmoudrik/deep-go-wrap

Repository files navigation

deep-go-wrap

Toolkit designed to ease development of your Deep Neural Network models for the game of Go.

HDF Go Datasets

  • different planes support, easy to extend
  • parallel processing of games
  • the HDF dataset created is compatible with pylearn2 for instance, but NOT with DeepCL. To create dataset for DeepCL, see hdf2deepcl_v2.py tool.

A naive bash example how to make a dataset

DATA_DIR="/path/to/dir/containing/your/games"

# make pseudorandomly shuffled list of all games found
find "$DATA_DIR" -name '*.sgf' | sort -R > filelist

# this creates the dataset, with 1 bit per plane (7 of them)
# for each goban point. The dataset is transparently gzip
# compressed by hdf5, so the size is managable.
cat filelist | ./process_sgf.py -p clark_storkey_2014_packed dataset.hdf5

Comparison of different dataset making options

The following list summarizes file size for different options. The summary was made from 200 random GoGoD games (39805 example pairs ~ 200 pairs per game). Running times were basically the same (~ 1.8 sec per game on commodity laptop). HDF5 compresses the dataset transparently using gzip -9. The feature cube is composed of 7 planes used in Clark an Storkey's 2014 paper. The --flatten option just reshapes data from (7,19,19) to (7*19*19,).

  • Fully expanded, flattened data in floats32 (e.g. directly ready for pylearn2)
    • 14.2kB per game: --dtype float32 --flatten -l expanded_label -p clark_storkey_2014
  • Fully expanded, flattened data in uint8
    • 7.7kB per game: --flatten -l expanded_label -p clark_storkey_2014
  • Flattened data, both features and labels packed using numpy.packbits
    • 3.2kB per game: --flatten -l expanded_label_packed -p clark_storkey_2014_packed
  • Flattened data, features packed using numpy.packbits, label just one class number
    • 2.9kB per game: --flatten -l simple_label -p clark_storkey_2014_packed

GTP wrapper

  • interface for plugging in various Deep Network architectures
    • caffe nets, DeepCL nets (working), easy to extend
  • I/O handling, data planes extraction
  • full GTP support using gomill library
  • pass/resign implementation, using GnuGo as an oracle. Note that this slows thigs down a bit, GnuGo being slower than CNN.
  • move correctness checking
  • Do you have other great ideas? Contribute, or make an issue!

caffe network setup

  1. Adding up new net is easy. Here I use Detlef Schmicker's model as an example -- see deepgo/bot_caffe.py.

  2. You can get (Detlef's 54% CNN)[http://physik.de/CNNlast.tar.gz]

  3. Edit path to prototxt and trained model file in the deepgowrap.py (you may have to uncomment main_deepcl in the code)

    • now the deepgowrap acts as an GTP program, so you can run it:

    echo -e "boardsize 19\nclear_board B\nquit" | ./deepgowrap.py

DeepCL setup (slightly obsolete)

  1. Train your deepcl model

    1. get data https://github.com/hughperkins/kgsgo-dataset-preprocessor

    2. get DeepCL, probably my fork of Hugh Perkins's repo which is sure to work, the changes aren't big and will probably be merged to the Hugh's repo, so this is for the time being.

    3. train your model

    4. or, instead of the last step, just download my test small weights (precision 19.6%, trained in 10 minutes, extremely weak, from here (binary file, might not work on different archs), trained with:

      ./deepclrun numtrain=200000 dataset=kgsgoall 'netdef=32c3{z}-relu-4*(8c3{z}-relu)-128n-tanh-361n' numepochs=20 learningrate=0.0001

  2. edit path to DeepCL and model's weights file in the deepgowrap.py (you may have to uncomment main_deepcl in the code)

    • now the deepgowrap acts as an GTP program, so you can run it:

    echo -e "boardsize 19\nclear_board B\nquit" | ./deepgowrap.py

Requirements

About

Toolkit designed to ease development of your Deep Neural Network models for the game of Go (weiqi, baduk).

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages