b32a4724dd2f8c97ff3b3813e5622d6bb6fb8f7a,dual_net.py,DualNetwork,run_many,#DualNetwork#Any#Any#,231

Before Change

        return probabilities[0], value[0]

    def run_many(self, positions, use_random_symmetry=True):
        fts = functools.partial(features.extract_features, features=features.NEW_FEATURES)
        processed = list(map(fts, positions))
        if use_random_symmetry:
            syms_used, processed = features.randomize_symmetries_feat(processed)
        probabilities, value = self.session.run(
            [self.policy_output, self.value_output],
            feed_dict={self.x: processed,
                       self.train_mode: False})
        if use_random_symmetry:
            probabilities = features.invert_symmetries_pi(syms_used, probabilities)
        return probabilities, value

After Change

        return probs[0], values[0]

    def run_many(self, positions, use_random_symmetry=True):
        processed = list(map(features.extract_features, positions))
        if use_random_symmetry:
            syms_used, processed = symmetries.randomize_symmetries_feat(processed)
        outputs = self.sess.run(self.inference_output,
            feed_dict={self.inference_input["pos_tensor"]: processed})
        probabilities, value = outputs["policy_output"], outputs["value_output"]
        if use_random_symmetry:
            probabilities = symmetries.invert_symmetries_pi(syms_used, probabilities)
        return probabilities, value

Instances