From 55898d0e66db9e902acb336d5816e343a7d2f94d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christoffer=20M=C3=BCller=20Madsen?=
 <christoffer@guava.space>
Date: Mon, 12 Mar 2018 00:11:55 +0100
Subject: [PATCH] renaming parameters

---
 README.org |  6 +++---
 main.py    | 14 ++++++++++----
 network.py |  4 ++--
 3 files changed, 15 insertions(+), 9 deletions(-)

diff --git a/README.org b/README.org
index 61549cc..2fed69a 100644
--- a/README.org
+++ b/README.org
@@ -56,7 +56,7 @@ The following examples describe commmon operations.
 
 *** Evaluate model named =quack= using default evaluation method (currently =random=)
 
-=python3 --eval --model-name=quack=
+=python3 --eval --model=quack=
 
 *** Evaluate default model using evaluation methods =random= and =pubeval=
 
@@ -73,14 +73,14 @@ directory. Otherwise, the model is stored in =models/$MODEL=.
 Along with the Tensorflow checkpoint files in the directory, the following files
 are stored:
 
-- =model.episodes=: The number of episodes of training performed with the
+- =episodes_trained=: The number of episodes of training performed with the
   model
 - =logs/eval.log=: Log of all completed evaluations performed on the model. The
   format of this file is specified in [[Log format]].
 - =logs/train.log=: Log of all completed training sessions performed on the
   model. If a training session is aborted before the pre-specified episode
   target is reached, nothing will be written to this file, although
-  =model.episodes= will be updated every time the model is saved to disk. The
+  =episodes_trained= will be updated every time the model is saved to disk. The
   format of this file is specified in [[Log format]].
 
 ** Log format
diff --git a/main.py b/main.py
index 0bfcea3..2012b0c 100644
--- a/main.py
+++ b/main.py
@@ -3,6 +3,12 @@ import sys
 import os
 import time
 
+models_storage_path = 'models'
+
+# Create models folder
+if not os.path.exists(models_storage_path):
+    os.makedirs(models_storage_path)
+
 # Define helper functions
 def log_train_outcome(outcome, trained_eps = 0):
     format_vars = { 'trained_eps': trained_eps,
@@ -34,9 +40,9 @@ parser = argparse.ArgumentParser(description="Backgammon games")
 parser.add_argument('--episodes', action='store', dest='episode_count',
                     type=int, default=1000,
                     help='number of episodes to train')
-parser.add_argument('--model-path', action='store', dest='model_path',
-                    default='./model',
-                    help='path to Tensorflow model')
+parser.add_argument('--model', action='store', dest='model',
+                    default='default',
+                    help='name of Tensorflow model to use')
 parser.add_argument('--eval-methods', action='store',
                     default=['random'], nargs='*',
                     help='specifies evaluation methods')
@@ -55,7 +61,7 @@ parser.add_argument('--start-episode', action='store', dest='start_episode',
 args = parser.parse_args()
 
 config = {
-    'model_path': args.model_path,
+    'model_path': os.path.join(models_storage_path, args.model),
     'episode_count': args.episode_count,
     'eval_methods': args.eval_methods,
     'train': args.train,
diff --git a/network.py b/network.py
index 435ccde..21da4a7 100644
--- a/network.py
+++ b/network.py
@@ -26,7 +26,7 @@ class Network:
 
 
         # Restore trained episode count for model
-        episode_count_path = os.path.join(self.checkpoint_path, "model.episodes")
+        episode_count_path = os.path.join(self.checkpoint_path, "episodes_trained")
         if os.path.isfile(episode_count_path):
             with open(episode_count_path, 'r') as f:
                 self.config['start_episode'] = int(f.read())
@@ -91,7 +91,7 @@ class Network:
 
     def save_model(self, episode_count):
         self.saver.save(self.session, os.path.join(self.checkpoint_path, 'model.ckpt'))
-        with open(os.path.join(self.checkpoint_path, "model.episodes"), 'w+') as f:
+        with open(os.path.join(self.checkpoint_path, "episodes_trained"), 'w+') as f:
             f.write(str(episode_count) + "\n")
     
     def restore_model(self):