initial commit

Author: carpedm20

.gitignore

@@ -0,0 +1,124 @@
+# Data
+data/hand
+data/gaze
+data/*.png
+samples
+outputs
+
+# Log
+logs
+
+# ETC
+paper.pdf
+
+# Created by https://www.gitignore.io/api/python,vim
+
+### Python ###
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*,cover
+.hypothesis/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# dotenv
+.env
+
+# virtualenv
+.venv/
+venv/
+ENV/
+
+# Spyder project settings
+.spyderproject
+
+# Rope project settings
+.ropeproject
+
+
+### Vim ###
+# swap
+[._]*.s[a-v][a-z]
+[._]*.sw[a-p]
+[._]s[a-v][a-z]
+[._]sw[a-p]
+# session
+Session.vim
+# temporary
+.netrwhist
+*~
+# auto-generated tag files
+tags
+
+# End of https://www.gitignore.io/api/python,vim

README.md

@@ -0,0 +1,33 @@
+# Neural Combinatorial Optimization in Tensorflow
+
+TensorFlow implementation of [Neural Combinatorial Optimization with Reinforcement Learning](http://arxiv.org/abs/1611.09940).
+
+![model](./assets/model.png)
+
+
+## Requirements
+
+- Python 2.7
+- [TensorFlow](https://www.tensorflow.org/) 0.12.0
+- [tqdm](https://github.com/tqdm/tqdm)
+
+
+## Usage
+
+To train a model:
+
+    $ python main.py
+    $ tensorboard --logdir=logs --host=0.0.0.0
+
+To test a model:
+
+    $ python main.py --is_train=False
+
+## Results
+
+(in progress)
+
+
+## Author
+
+Taehoon Kim / [@carpedm20](http://carpedm20.github.io)

assets/model.png

@@ -0,0 +1,52 @@
+#-*- coding: utf-8 -*-
+import argparse
+
+def str2bool(v):
+  return v.lower() in ('true', '1')
+
+arg_lists = []
+parser = argparse.ArgumentParser()
+
+def add_argument_group(name):
+  arg = parser.add_argument_group(name)
+  arg_lists.append(arg)
+  return arg
+
+# Network
+net_arg = add_argument_group('Network')
+net_arg.add_argument('--hidden_dims', type=int, default=200, help='')
+net_arg.add_argument('--num_layer', type=int, default=2, help='')
+
+# Data
+data_arg = add_argument_group('Data')
+data_arg.add_argument('--data_set', type=str, default='gaze')
+data_arg.add_argument('--data_dir', type=str, default='data')
+data_arg.add_argument('--input_height', type=int, default=35)
+data_arg.add_argument('--input_width', type=int, default=55)
+data_arg.add_argument('--task_name', type=str, default='TPS20')
+
+# Training / test parameters
+train_arg = add_argument_group('Training')
+train_arg.add_argument('--is_train', type=str2bool, default=True, help='')
+train_arg.add_argument('--optimizer', type=str, default='rmsprop', help='')
+train_arg.add_argument('--max_step', type=int, default=10000, help='')
+train_arg.add_argument('--reg_scale', type=float, default=0.5, help='')
+train_arg.add_argument('--batch_size', type=int, default=512, help='')
+train_arg.add_argument('--learning_rate', type=float, default=0.001, help='')
+train_arg.add_argument('--checkpoint_secs', type=int, default=300, help='')
+train_arg.add_argument('--max_grad_norm', type=float, default=50, help='')
+
+# Misc
+misc_arg = add_argument_group('Misc')
+misc_arg.add_argument('--log_step', type=int, default=20, help='')
+misc_arg.add_argument('--log_dir', type=str, default='logs')
+misc_arg.add_argument('--sample_dir', type=str, default='samples')
+misc_arg.add_argument('--output_dir', type=str, default='outputs')
+misc_arg.add_argument('--load_path', type=str, default='')
+misc_arg.add_argument('--debug', type=str2bool, default=False)
+misc_arg.add_argument('--gpu_memory_fraction', type=float, default=1.0)
+misc_arg.add_argument('--random_seed', type=int, default=123, help='')
+
+def get_config():
+  config, unparsed = parser.parse_known_args()
+  return config, unparsed

config.py

@@ -0,0 +1,3 @@
+class DataLoader(object):
+  def __init__(self):
+    pass

data_loader.py

@@ -0,0 +1,29 @@
+import sys
+import numpy as np
+import tensorflow as tf
+
+from trainer import Trainer
+from config import get_config
+from utils import prepare_dirs, save_config
+
+config = None
+
+def main(_):
+  prepare_dirs(config)
+
+  rng = np.random.RandomState(config.random_seed)
+  tf.set_random_seed(config.random_seed)
+
+  trainer = Trainer(config, rng)
+  save_config(config.model_dir, config)
+
+  if config.is_train:
+    trainer.train()
+  else:
+    if not config.load_path:
+      raise Exception("[!] You should specify `load_path` to load a pretrained model")
+    trainer.test()
+
+if __name__ == "__main__":
+  config, unparsed = get_config()
+  tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)

layers.py

@@ -0,0 +1,59 @@
+import tensorflow as tf
+from tensorflow.contrib.framework import arg_scope
+
+from layers import *
+from utils import show_all_variables
+
+class Model(object):
+  def __init__(self, config, data_loader):
+    self.data_loader = data_loader
+
+    self.task = config.task
+    self.debug = config.debug
+    self.config = config
+
+    self.input_height = config.input_height
+    self.input_width = config.input_width
+    self.input_channel = config.input_channel
+
+    self.reg_scale = config.reg_scale
+    self.learning_rate = config.learning_rate
+    self.max_grad_norm = config.max_grad_norm
+    self.batch_size = config.batch_size
+
+    self.layer_dict = {}
+
+    self._build_placeholders()
+    self._build_model()
+    self._build_optim()
+
+    show_all_variables()
+
+  def _build_placeholders(self):
+    self.inputs = tf.placeholder(tf.float32, name="inputs")
+    self.lengths = tf.placeholder(tf.float32, name="lengths")
+    self.targets = tf.placeholder(tf.float32, name="targets")
+    self.is_train = tf.placeholder(tf.bool, name="is_train")
+
+  def _build_encoder(self):
+    self.global_step = tf.Variable(0, trainable=False)
+
+    with tf.variable_scope("encoder"):
+      self.cell = tf.nn.rnn_cell.LSTMCell(size)
+      if num_layers > 1:
+        self.cell = tf.nn.rnn_cell.MultiRNNCell([single_cell] * num_layers)
+
+      self.rnn = tf.nn.dynamic_rnn(self.cell, self.inputs, self.seq_length)
+
+    with tf.variable_scope("dencoder"):
+      self.cell = tf.nn.rnn_cell.LSTMCell(size)
+      if num_layers > 1:
+        self.cell = tf.nn.rnn_cell.MultiRNNCell([single_cell] * num_layers)
+
+      self.rnn = tf.nn.dynamic_rnn(self.cell, self.inputs, self.seq_length)
+
+  def _build_optim(self):
+    self.loss = tf.reduce_mean(self.output - self.targets)
+
+    self.learning_rate = tf.Variable(self.learning_rate)
+    self.optim = tf.train.AdamOptimizer(self.learning_rate)

main.py

@@ -0,0 +1,71 @@
+import os
+import numpy as np
+from tqdm import trange
+import tensorflow as tf
+from tensorflow.contrib.framework.python.ops import arg_scope
+
+from model import Model
+from data_loader import DataLoader
+
+class Trainer(object):
+  def __init__(self, config, rng):
+    self.config = config
+    self.rng = rng
+
+    self.task = config.task
+    self.model_dir = config.model_dir
+    self.gpu_memory_fraction = config.gpu_memory_fraction
+
+    self.log_step = config.log_step
+    self.max_step = config.max_step
+    self.checkpoint_secs = config.checkpoint_secs
+
+    self.summary_ops = {}
+
+    self.model = Model(config, self.data_loader)
+    self.data_loader = DataLoader(config, rng=self.rng)
+
+    self._build_session()
+
+  def _build_session(self):
+    self.saver = tf.train.Saver()
+    self.summary_writer = tf.summary.FileWriter(self.model_dir)
+
+    sv = tf.train.Supervisor(logdir=self.model_dir,
+                             is_chief=True,
+                             saver=self.saver,
+                             summary_op=None,
+                             summary_writer=self.summary_writer,
+                             save_summaries_secs=300,
+                             save_model_secs=self.checkpoint_secs,
+                             global_step=self.model.discrim_step)
+
+    gpu_options = tf.GPUOptions(
+        per_process_gpu_memory_fraction=self.gpu_memory_fraction,
+        allow_growth=True) # seems to be not working
+    sess_config = tf.ConfigProto(allow_soft_placement=True,
+                                 gpu_options=gpu_options)
+
+    self.sess = sv.prepare_or_wait_for_session(config=sess_config)
+
+  def train(self):
+    print("[*] Training starts...")
+    pass
+
+  def test(self):
+    pass
+
+  def _inject_summary(self, tag, feed_dict, step):
+    summaries = self.sess.run(self.summary_ops[tag], feed_dict)
+    self.summary_writer.add_summary(summaries['summary'], step)
+
+    path = os.path.join(
+        self.config.sample_model_dir, "{}.png".format(step))
+    imwrite(path, img_tile(summaries['output'],
+            tile_shape=self.config.sample_image_grid)[:,:,0])
+
+  def _get_summary_writer(self, result):
+    if result['step'] % self.log_step == 0:
+      return self.summary_writer
+    else:
+      return None