DouZero icon indicating copy to clipboard operation
DouZero copied to clipboard

[ICML 2021] DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning | 斗地主AI

Results 32 DouZero issues
Sort by recently updated
recently updated
newest added

这是我的参数设置 "args": { "actor_device_cpu": false, "alpha": 0.99, "batch_size": 32, "disable_checkpoint": false, "epsilon": 1e-05, "exp_epsilon": 0.01, "gpu_devices": "0,1,2,3,4", "learning_rate": 0.0001, "load_model": false, "max_grad_norm": 40.0, "momentum": 0, "num_actor_devices": 4, "num_actors": 100, "num_buffers":...

I don't think "Suits are irrelevant in DouDizhu". For example, the three landlord cards may contain a spade 8, while landlord could hold another diamond 8. If the landlord plays...

actor网络在收集数据时使用了model前向传播,leaner对网络参数在进行跟新,二者访问同一共享内存是否会出现冲突,但是actor中并没有用锁?

Traceback (most recent call last): File "/content/drive/MyDrive/DouZero/evaluate.py", line 24, in evaluate(args.landlord, File "/content/drive/MyDrive/DouZero/douzero/evaluation/simulation.py", line 48, in evaluate card_play_data_list = pickle.load(f) EOFError: Ran out of input 在使用python3 evaluate.py --landlord /baselines/sl/landlord.ckpt --landlord_up...

![image](https://user-images.githubusercontent.com/51454565/234094382-b7b62807-3ad5-4bb2-b9d1-900c08c2e45f.png)

![image](https://user-images.githubusercontent.com/53806323/205202458-e39896e2-9f20-44e7-97ad-92471458e4dd.png) 胜率很低 有什么办法增强网络吗

有什么办法降低ram使用吗?不然根本没办法跑

在Windows系统上使用GPU进行模拟时,遇到 `operation not supported` 错误通常是由于Windows不支持CUDA tensor上的多进程操作。这是一个已知问题,因为CUDA多进程支持在Windows上的实现存在一些限制。然而,有一些解决方案可以帮助解决或绕过这个问题。 ### 解决方案 #### 1. 使用单进程来处理CUDA操作 如果多进程是为了并行计算,可以考虑在单个进程中使用并行库,如CUDA Streams或PyTorch的DataLoader中的多线程模式。这可以避免多进程的限制。 ```python import torch from torch.utils.data import DataLoader # 使用DataLoader的多线程模式 data_loader = DataLoader(dataset, batch_size=64, shuffle=True, num_workers=4, pin_memory=True) for...