Atari100k

该数据集是一个被广泛用于评估强化学习算法样本效率的基准，它将算法与环境的交互限制在10万步以内（考虑到4步跳过，相当于400,000帧）。此外，该数据集用于在26款Atari游戏中评估各种强化学习模型的性能。其规模为10万次交互，任务是对强化学习算法的样本效率进行评估。

This dataset is a widely adopted benchmark for evaluating the sample efficiency of reinforcement learning algorithms. It restricts the interaction between the algorithm and the environment to 100,000 steps, which equals 400,000 frames when taking the 4-frame skip into consideration. Additionally, this dataset is utilized to evaluate the performance of diverse reinforcement learning models across 26 Atari games. With a scale of 100,000 interactions, its primary objective is to assess the sample efficiency of reinforcement learning algorithms.

搜集汇总

背景与挑战

背景概述

该数据集实际上是一个名为Drama的Mamba/Mamba2驱动的模型基于强化学习的代理项目，提供了训练和评估的详细指令，并引用了相关论文和代码参考。