Parallel example

Coach.py

@@ -1,12 +1,12 @@
 import logging
 import os
 import sys
-from collections import deque
+from multiprocessing import Manager
+from multiprocessing import Process
 from pickle import Pickler, Unpickler
 from random import shuffle
 
 import numpy as np
-from tqdm import tqdm
 
 from Arena import Arena
 from MCTS import MCTS
@@ -25,11 +25,10 @@ def __init__(self, game, nnet, args):
         self.nnet = nnet
         self.pnet = self.nnet.__class__(self.game)  # the competitor network
         self.args = args
-        self.mcts = MCTS(self.game, self.nnet, self.args)
         self.trainExamplesHistory = []  # history of examples from args.numItersForTrainExamplesHistory latest iterations
         self.skipFirstSelfPlay = False  # can be overriden in loadTrainExamples()
 
-    def executeEpisode(self):
+    def executeEpisode(self, return_examples=None):
         """
         This function executes one episode of self-play, starting with player 1.
         As the game is played, each turn is added as a training example to
@@ -50,12 +49,13 @@ def executeEpisode(self):
         self.curPlayer = 1
         episodeStep = 0
 
+        mcts = MCTS(self.game, self.nnet, self.args)
         while True:
             episodeStep += 1
             canonicalBoard = self.game.getCanonicalForm(board, self.curPlayer)
             temp = int(episodeStep < self.args.tempThreshold)
 
-            pi = self.mcts.getActionProb(canonicalBoard, temp=temp)
+            pi = mcts.getActionProb(canonicalBoard, temp=temp)
             sym = self.game.getSymmetries(canonicalBoard, pi)
             for b, p in sym:
                 trainExamples.append([b, self.curPlayer, p, None])
@@ -66,7 +66,8 @@ def executeEpisode(self):
             r = self.game.getGameEnded(board, self.curPlayer)
 
             if r != 0:
-                return [(x[0], x[2], r * ((-1) ** (x[1] != self.curPlayer))) for x in trainExamples]
+                return_examples += [(x[0], x[2], r * ((-1) ** (x[1] != self.curPlayer))) for x in trainExamples]
+                return
 
     def learn(self):
         """
@@ -80,23 +81,27 @@ def learn(self):
         for i in range(1, self.args.numIters + 1):
             # bookkeeping
             log.info(f'Starting Iter #{i} ...')
+
             # examples of the iteration
             if not self.skipFirstSelfPlay or i > 1:
-                iterationTrainExamples = deque([], maxlen=self.args.maxlenOfQueue)
-
-                for _ in tqdm(range(self.args.numEps), desc="Self Play"):
-                    self.mcts = MCTS(self.game, self.nnet, self.args)  # reset search tree
-                    iterationTrainExamples += self.executeEpisode()
-
-                # save the iteration examples to the history 
-                self.trainExamplesHistory.append(iterationTrainExamples)
+                manager = Manager()
+                examples = manager.list()
+                processes = []
+                for _ in range(self.args.numEps):
+                    p = Process(target=self.executeEpisode, args=(examples,))
+                    p.start()
+                    processes.append(p)
+                [p.join() for p in processes]
+
+                # save the iteration examples to the history
+                self.trainExamplesHistory.append(examples[-self.args.maxlenOfQueue:])
 
             if len(self.trainExamplesHistory) > self.args.numItersForTrainExamplesHistory:
                 log.warning(
                     f"Removing the oldest entry in trainExamples. len(trainExamplesHistory) = {len(self.trainExamplesHistory)}")
                 self.trainExamplesHistory.pop(0)
             # backup history to a file
-            # NB! the examples were collected using the model from the previous iteration, so (i-1)  
+            # NB! the examples were collected using the model from the previous iteration, so (i-1)
             self.saveTrainExamples(i - 1)
 
             # shuffle examples before training

utils.py

@@ -19,4 +19,4 @@ def update(self, val, n=1):
 
 class dotdict(dict):
     def __getattr__(self, name):
-        return self[name]
+        return self.get(name)