ab0f3afc45ae2b53fee03be82f4cbc037f874222,ml/rl/evaluation/policy_evaluator.py,PolicyEvaluator,evaluate_slates,#PolicyEvaluator#Any#Any#Any#Any#,37

Before Change

                action_probabilities[action_selection[slate_index]]
            ips_denominator = baseline_probabilities[slate_index]
            value_sum += ips_numerator / ips_denominator
            logger.debug(
                "DETAILS",
                rewards[slate_index],
                action_probabilities[action_selection[slate_index]],
                ips_denominator,
                value_sum,
            )
        return value_sum / num_examples

After Change

        num_slates = len(slates)
        value_sum = 0.0
        for slate_index, slate in enumerate(slates):
            for name, value in slate.policy_net_features.items():
                workspace.FeedBlob(name, value)
            for model_name, model_input in slate.model_feature_set.items():
                value_input_model_params = self.value_input_models[model_name]
                print(model_input)
                for name, value in model_input.items():

Instances