hongshaorou/Tianshou
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

minor fix in core/policy

Browse Source
This commit is contained in:
haosheng 2017-12-11 13:25:22 +08:00
parent e3c0478fa0
commit 88ecaa332d
2 changed files with 13 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
tianshou/core/policy/base.py
View File

@ -148,19 +148,19 @@ class StochasticPolicy(object):
""" """
return self._act(observation) return self._act(observation)
if n_samples is None: # if n_samples is None:
samples = self._sample(n_samples=1) # samples = self._sample(n_samples=1)
return tf.squeeze(samples, axis=0) # return tf.squeeze(samples, axis=0)
elif isinstance(n_samples, int): # elif isinstance(n_samples, int):
return self._sample(n_samples) # return self._sample(n_samples)
else: # else:
n_samples = tf.convert_to_tensor(n_samples, dtype=tf.int32) # n_samples = tf.convert_to_tensor(n_samples, dtype=tf.int32)
_assert_rank_op = tf.assert_rank( # _assert_rank_op = tf.assert_rank(
n_samples, 0, # n_samples, 0,
message="n_samples should be a scalar (0-D Tensor).") # message="n_samples should be a scalar (0-D Tensor).")
with tf.control_dependencies([_assert_rank_op]): # with tf.control_dependencies([_assert_rank_op]):
samples = self._sample(n_samples) # samples = self._sample(n_samples)
return samples # return samples
def _act(self, observation): def _act(self, observation):
""" """

1
tianshou/core/policy/stochastic.py
View File

@ -76,7 +76,6 @@ class OnehotCategorical(StochasticPolicy):
return sampled_action return sampled_action
def _log_prob(self, sampled_action): def _log_prob(self, sampled_action):
sampled_action_onehot = tf.one_hot(sampled_action, self.n_categories, dtype=self.act_dtype)
return -tf.nn.sparse_softmax_cross_entropy_with_logits(labels=sampled_action, logits=self.logits) return -tf.nn.sparse_softmax_cross_entropy_with_logits(labels=sampled_action, logits=self.logits)
# given = tf.cast(given, self.param_dtype) # given = tf.cast(given, self.param_dtype)