param dumper

examples/Atari2600/DQN.py

@@ -115,9 +115,7 @@ class Model(ModelDesc):
 
         target = reward + (1.0 - tf.cast(isOver, tf.float32)) * GAMMA * tf.stop_gradient(best_v)
 
-        sqrcost = tf.square(target - pred_action_value)
-        abscost = tf.abs(target - pred_action_value)    # robust error func
-        cost = tf.select(abscost < 1, sqrcost, abscost)
+        cost = symbf.clipped_l2_loss(target - pred_action_value)
         summary.add_param_summary([('conv.*/W', ['histogram', 'rms']),
                                    ('fc.*/W', ['histogram', 'rms']) ])   # monitor all W
         self.cost = tf.reduce_mean(cost, name='cost')

examples/Atari2600/common.py

@@ -39,12 +39,21 @@ def eval_with_funcs(predict_funcs, nr_eval):
     class Worker(StoppableThread):
         def __init__(self, func, queue):
             super(Worker, self).__init__()
-            self.func = func
+            self._func = func
             self.q = queue
+
+        def func(self, *args, **kwargs):
+            if self.stopped():
+                raise RuntimeError("stopped!")
+            return self._func(*args, **kwargs)
+
         def run(self):
             player = get_player()
             while not self.stopped():
-                score = play_one_episode(player, self.func)
+                try:
+                    score = play_one_episode(player, self.func)
+                except RuntimeError:
+                    return
                 self.queue_put_stoppable(self.q, score)
 
     q = queue.Queue(maxsize=2)

scripts/dump-model-params.py

@@ -7,23 +7,39 @@ import argparse
 import tensorflow as tf
 import imp
 
-from tensorpack.utils import *
+from tensorpack import *
 from tensorpack.tfutils import sessinit, varmanip
-from tensorpack.dataflow import *
 
 parser = argparse.ArgumentParser()
-parser.add_argument(dest='config')
+parser.add_argument('--config', help='config file')
+parser.add_argument('--meta', help='metagraph file')
 parser.add_argument(dest='model')
 parser.add_argument(dest='output')
 args = parser.parse_args()
 
-get_config_func = imp.load_source('config_script', args.config).get_config
+assert args.config or args.meta, "Either config or metagraph must be present!"
 
 with tf.Graph().as_default() as G:
-    config = get_config_func()
-    config.model.build_graph(config.model.get_input_vars(), is_training=False)
+    if args.config:
+        MODEL = imp.load_source('config_script', args.config).Model
+        M = MODEL()
+        M.build_graph(M.get_input_vars(), is_training=False)
+    else:
+        M = ModelFromMetaGraph(args.meta)
+
+    # loading...
     init = sessinit.SaverRestore(args.model)
     sess = tf.Session()
     init.init(sess)
+
+    # dump ...
     with sess.as_default():
-        varmanip.dump_session_params(args.output)
+        if args.output.endswith('npy'):
+            varmanip.dump_session_params(args.output)
+        else:
+            var = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
+            var.extend(tf.get_collection(EXTRA_SAVE_VARS_KEY))
+            logger.info("Variables to dump:")
+            logger.info(", ".join([v.name for v in var]))
+            saver = tf.train.Saver(var_list=var)
+            saver.save(sess, args.output, write_meta_graph=False)

tensorpack/RL/gymenv.py

@@ -24,9 +24,9 @@ class GymEnv(RLEnvironment):
     """
     def __init__(self, name, dumpdir=None, viz=False):
         self.gymenv = gym.make(name)
-        #if dumpdir:
-            #mkdir_p(dumpdir)
-            #self.gymenv.monitor.start(dumpdir, force=True, seed=0)
+        if dumpdir:
+            mkdir_p(dumpdir)
+            self.gymenv.monitor.start(dumpdir)
 
         self.reset_stat()
         self.rwd_counter = StatCounter()

tensorpack/tfutils/symbolic_functions.py

@@ -78,6 +78,13 @@ def rms(x, name=None):
             return tf.sqrt(tf.reduce_mean(tf.square(x)), name=name)
     return tf.sqrt(tf.reduce_mean(tf.square(x)), name=name)
 
+def clipped_l2_loss(x, name=None):
+    if name is None:
+        name = 'clipped_l2_loss'
+    sqrcost = tf.square(x)
+    abscost = tf.abs(x)
+    return tf.select(abscost < 1, sqrcost, abscost, name=name)
+
 def get_scalar_var(name, init_value):
     return tf.get_variable(name, shape=[],
             initializer=tf.constant_initializer(init_value),

tensorpack/tfutils/varmanip.py

@@ -9,6 +9,7 @@ from collections import defaultdict
 import re
 import numpy as np
 from ..utils import logger
+from ..utils.naming import *
 
 __all__ = ['SessionUpdate', 'dump_session_params', 'dump_chkpt_vars',
         'get_savename_from_varname']