huawei-noah / xingtian

xingtian is a componentized library for the development and verification of reinforcement learning algorithms
MIT License
301 stars 89 forks source link

pbt_breakout_impala.yaml执行有问题,报错如图 #19

Open muranran opened 2 years ago

muranran commented 2 years ago

d2841c19888965f6db5fb362fdb7c5b

muranran commented 2 years ago

image 报错如上,但是可以正常运行 稳定后数据有问题 image reward一直为0

hustqj commented 2 years ago

把你运行的配置附上

muranran commented 2 years ago

配置如下 image image image image image

muranran commented 2 years ago

image

muranran commented 2 years ago

image

muranran commented 2 years ago

配置 image 在训练30min后,结果如下 image 训练数据train_avg_reward一直在0左右徘徊,没有报错

muranran commented 2 years ago

运行1h后报错 image

muranran commented 2 years ago

最终 image