Commits

Breadcrumbs

History for

onse-minigo

User selector

All users

All time

Commit history

Commits on Apr 3, 2018

• Clean up Python code.

  Show description for 2d2a6a2
  gcp

  committed
  2d2a6a2

  Copy full SHA for 2d2a6a2

• Support batches larger than GPU RAM size.

  Show description for 7f8ff14
  

  ywrt

  authored and

  gcp

  committed
  7f8ff14

  Copy full SHA for 7f8ff14

Commits on Mar 20, 2018

• Fix v2 pipeline: shuffle validation set.

  Show description for 8390055
  gcp

  committed
  8390055

  Copy full SHA for 8390055

Commits on Mar 14, 2018

• Add a unittest for the FileDataSrc class.

  Show description for 66eb43b
  

  ywrt

  authored and

  gcp

  committed
  66eb43b

  Copy full SHA for 66eb43b

• Bugfix FileDataSrc.

  Show description for 7a76eb5
  

  Error323

  authored and

  gcp

  committed
  7a76eb5

  Copy full SHA for 7a76eb5

Commits on Mar 7, 2018

• Fix training to work on Windows again.

  Show description for ee5f391
  

  ywrt

  authored and

  gcp

  committed
  ee5f391

  Copy full SHA for ee5f391

Commits on Mar 2, 2018

• Move ChunkParser into a seperate file.

  Show description for 3bf9a7f
  

  ywrt

  authored and

  gcp

  committed
  3bf9a7f

  Copy full SHA for 3bf9a7f

Commits on Feb 28, 2018

• Update TensorFlow pipeline.

  Show description for 800990c
  

  ywrt

  authored and

  gcp

  committed
  800990c

  Copy full SHA for 800990c

Commits on Jan 30, 2018

• Increase skip factor for chunk parsing.

  Show description for ef66e26
  gcp

  committed
  ef66e26

  Copy full SHA for ef66e26

• Speed up chunk parsing/data split.

  Show description for 9e9fd49
  gcp

  committed
  9e9fd49

  Copy full SHA for 9e9fd49

• Make net_to_model work again, print policy loss.

  Show description for 1559ada
  gcp

  committed
  1559ada

  Copy full SHA for 1559ada

• Use a feedable iterator rather than tf.cond().

  Show description for 012bd8c
  gcp

  committed
  012bd8c

  Copy full SHA for 012bd8c

• Let chunk parsing skip every 1/8 positions.

  Show description for b43f29f
  gcp

  committed
  b43f29f

  Copy full SHA for b43f29f

• Dump supervised learning per game. Shuffle chunks.

  Show description for d450e8e
  gcp

  committed
  d450e8e

  Copy full SHA for d450e8e

• Implement real validation split for TF.

  Show description for d396c6a
  gcp

  committed
  d396c6a

  Copy full SHA for d396c6a

Commits on Jan 23, 2018

• Update default TensorFlow parameters.

  Show description for cabe7d1
  gcp

  committed
  cabe7d1

  Copy full SHA for cabe7d1

Commits on Dec 26, 2017

• Add tests for TensorFlow parsing.

  Show description for cbf0e3a
  

  michael

  authored and

  gcp

  committed
  cbf0e3a

  Copy full SHA for cbf0e3a

• Make training/tf/parse fully GPU-bound.

  Show description for 5f4f981
  

  michael

  authored and

  gcp

  committed
  5f4f981

  Copy full SHA for 5f4f981

• Speed up record processing in TensorFlow training.

  Show description for 6b01841
  

  michael

  authored and

  gcp

  committed
  6b01841

  Copy full SHA for 6b01841

Commits on Dec 6, 2017

• Training: fix Python docstrings delimiter.

  

  akdtg

  authored and

  gcp

  committed
  314f648

  Copy full SHA for 314f648

Commits on Nov 13, 2017

• Ignore faulty training data.

  Show description for 4157d2c
  gcp

  committed
  4157d2c

  Copy full SHA for 4157d2c

Commits on Nov 9, 2017

• Fix generation of symmetries.

  Show description for 8c2b257
  gcp

  committed
  8c2b257

  Copy full SHA for 8c2b257

Commits on Nov 8, 2017

• Use more multiprocessing workers, smaller prefetch.

  Show description for 4cd0d6c
  gcp

  committed
  4cd0d6c

  Copy full SHA for 4cd0d6c

• Add speed info to training, avoid excess precision.

  Show description for e9c1dad
  gcp

  committed
  e9c1dad

  Copy full SHA for e9c1dad

• Make TensorFlow queue construction multiprocessed.

  Show description for 9001902
  gcp

  committed
  9001902

  Copy full SHA for 9001902

• Use Datasets instead of feed_dict.

  Show description for c16a056
  gcp

  committed
  c16a056

  Copy full SHA for c16a056

• Add license to TensorFlow sources.

  Show description for c06b9cb
  gcp

  committed
  c06b9cb

  Copy full SHA for c06b9cb

• Augment training data with symmetries.

  Show description for 26322af
  gcp

  committed
  26322af

  Copy full SHA for 26322af

Commits on Nov 7, 2017

• Small cleanups and optimizations in TensorFlow.

  Show description for 3bb06a1
  gcp

  committed
  3bb06a1

  Copy full SHA for 3bb06a1

Commits on Nov 5, 2017

• Add support for restoring a model.

  Show description for 7637ccd
  gcp

  committed
  7637ccd

  Copy full SHA for 7637ccd

• Train value network.

  Show description for b7b393d
  gcp

  committed
  b7b393d

  Copy full SHA for b7b393d

• Update TensforFlow training code.

  Show description for 49b91b0
  gcp

  committed
  49b91b0

  Copy full SHA for 49b91b0

• Partial training support for TensorFlow.

  Show description for 15fddf8
  gcp

  committed
  15fddf8

  Copy full SHA for 15fddf8

Loading