kwai/DouZero

Issues

• 策略调整可行性

  #53 opened 2 years ago by SvenNJ
  3

• 模型输入一定需要other_hand_cards吗？

  #84 opened 2 months ago by thunder95
  0

• 能否分享一下sl训练的代码？ 不需要分享人类的训练数据！

  #82 opened 2 months ago by AlphaYe
  1

• 卡牌编码问题

  #83 opened 2 months ago by lucky666liguan
  0

• 在Windows系统上使用GPU进行模拟时，遇到 operation not supported 错误

  #81 opened 5 months ago by linkerlin
  0

• [readme suggestion] A gif is not good for demonstration, an embedded video would be much better

  #80 opened 5 months ago by zzz6519003
  0

• 另外，训练过程中注意CPU和显存 我是actor用CPU，training用GPU，CPU一定要留余地，同时thread不要加太高 否则容易爆显存

  #66 opened 2 years ago by FYNIXqwq
  1

• C#可以直接调用训练好后的模型吗？

  #79 opened 9 months ago by oookidoo
  0

• what's the output dimension of actor network?

  #78 opened 9 months ago by zkengz
  0

• 有人大佬实现了癞子玩法了吗？

  #73 opened a year ago by lixiaozhui
  1

• 有没有大佬有学习好的斗地主的模型?

  #77 opened 10 months ago by QQ35081215
  0

• 我自己train了一下，为什么出来的结果不像你们训练出的那样智能？

  #42 opened 3 years ago by profection
  15

• 请问如何使用baseline进行辅助训练，而不是单纯的evaluate

  #76 opened a year ago by cxk555
  0

• 一点想法 关于不洗牌模式的训练

  #65 opened 2 years ago by FYNIXqwq
  9

• 为什么同样是4个GPU我的训练时候的FPS很低呢，基本都在2000左右

  #49 opened 2 years ago by mfxiaosheng
  7

• 当我使用5张A100 80G训练时，如何提高训练速度或者应该如何修改符合机器的参数

  #74 opened a year ago by Cyclones-Y
  2

• 如何查是否过拟合或者欠拟合呢

  #75 opened a year ago by XliOK
  0

• Consideration of suit of 3 landlord cards

  #72 opened a year ago by zizhang-qiu
  0

• 想询问一下baseline里面的ai训练了多久，我训练了几个礼拜了，表现感觉不如baseline

  #71 opened a year ago by GGG3313
  2

• actor网络在收集数据时使用了model前向传播，leaner对网络参数再进行跟新，这两者对同一共享内存更新是否会有冲突？

  #67 opened 2 years ago by mengxi7
  1

• 勾吧pretrain models都没有 一训练全是0

  #68 opened 2 years ago by SchoolBug
  2

• EOFError: Ran out of input

  #70 opened a year ago by lingyunQAQ
  2

• 在colab在跑训练，时间到了可以从上一次继续么

  #69 opened a year ago by lingyunQAQ
  1

• 欢乐斗地主实测胜率很低 算法还有升级的办法吗？

  #62 opened 2 years ago by xinyujituan
  2

• 关于深度学习与强化学习

  #64 opened 2 years ago by Zz-Quan
  2

• Apple M1芯片支持模拟和训练吗

  #58 opened 2 years ago by zhaowenzhi
  1

• colab上的训练代码一运行，ram就炸了

  #60 opened 2 years ago by jdy18
  3

• 没有找到main文件

  #47 opened 2 years ago by amadeus11111
  2

• 多显卡瓶颈在第0卡上

  #57 opened 2 years ago by SvenNJ
  0

• 咨询下LSTM策略调整

  #56 opened 2 years ago by SvenNJ
  0

• 关于models.py 几个问题

  #50 opened 2 years ago by SvenNJ
  7

• 执行loss.backward()几个小时报错out of memory

  #54 opened 2 years ago by SvenNJ
  7

• 策略可能性

  #55 opened 2 years ago by SvenNJ
  0

• 抢地主功能什么时候开放

  #51 opened 2 years ago by SvenNJ
  1

• 求助，用CPU无法训练问题！！！

  #52 opened 2 years ago by BESTTENG
  1

• 显存很快就满了

  #48 opened 2 years ago by Julong-Zheng
  8

• 新手求问free_queue和full_queue分别是什么意思？这个就是actor和learner之间的缓存区吗？

  #45 opened 2 years ago by czqaq
  1

• AI会把大小王作为四带二的二带出去，这是明显掉胜率的走法

  #43 opened 3 years ago by changjiuxiong
  4

• 关于evaluation阶段

  #46 opened 3 years ago by AlexWanghaoming
  1

• can you share the SL training code?

  #41 opened 3 years ago by jokerlc
  6

• 使用人类数据训练SL模型

  #44 opened 3 years ago by ctycode
  3

• i cannot train by cuda 11.3 win10 gtx1060,how can i do？

  #38 opened 3 years ago by qq782645210
  2

• 请教下如果是其它规则的斗地主要如果修改？

  #40 opened 3 years ago by JakeWoki
  2

• C++ implementation

  #37 opened 3 years ago by yffbit
  1

• Would the bidding phase be open sourced?

  #36 opened 3 years ago by weichaoguo
  2

• 出牌未校验是否大过上家

  #34 opened 3 years ago by changjiuxiong
  3

• train.py 运行 After 0 (L:0 U:0 D:0) frames: @ 0.0 fps (avg@ 0.0 fps) (L:0.0 U:0.0 D:0.0)

  #29 opened 3 years ago by Roywaller
  4

• 有没有必要设置size变量

  #32 opened 3 years ago by yffbit
  3

• Why is the first GPU over-consumed?

  #31 opened 3 years ago by xlnwel
  4

• 配置需求

  #30 opened 3 years ago by aacai
  1