xiaoou 4e6ad234da feat: Neural-only training, experience replay, per-round eval ...

3个训练效率提升:
1. Neural filter: V训练只取player=2数据, 排除ISMCTS状态干扰
   旧格式(无player列)自动fallback全量数据
2. 经验回放: 保留最近3轮数据混入当前轮训练, 扩大分布覆盖
   存储到 paodekuai-trainer/data/replay_buffer.pkl
3. 每轮eval: 训练后跑200局--eval测胜率, 下降>10%自动回滚
   回滚到 model_prev.pt, 防止坏模型污染下轮数据

2026-07-13 11:26:40 +08:00

base

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

CloudAPI

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

Config/Code

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

Core

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

dll

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

dll_new

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

docs

chore: 清理冗余 + 文档更新至v3.1 + 打包清单

2026-07-07 14:56:43 +08:00

exe

feat: 训练效率优化 — B:全bot记录 + A:并行调度

2026-07-13 03:25:51 +08:00

GameDAL

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

GameFix

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

GameModule

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

GameNetModule

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

GlobalSever

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

hjha-console

feat: --eval 评估模式 (ISMCTS/ISMCTS/NEURAL, 不生成CSV)

2026-07-13 08:38:18 +08:00

MrWu

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

NetWorkMessage

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

ObjectModel

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

PdkFriendServer

feat: CSV加player列+#RESULT汇总 — 训练中直接看胜率进化

2026-07-13 08:40:14 +08:00

scripts

feat: Neural-only training, experience replay, per-round eval

2026-07-13 11:26:40 +08:00

ServerCore

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

ServerData

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

training_data

feat: 远程Mac worker支持 — SSH分发训练任务

2026-07-13 03:29:23 +08:00

zyxAdapter

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

.editorconfig

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

.gitignore

chore: ignore bot-package zip artifacts

2026-07-07 15:36:03 +08:00

C:\Users\Administrator\Desktop\temp\gamepack.txt

fix: Windows paths → ./gamepack/, empty Console.ReadLine guard

2026-07-07 12:56:39 +08:00

C:\Users\Administrator\Desktop\temp\test\gamepack2026-07-07-12-47-41-559.txt

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

C:\Users\Administrator\Desktop\temp\test\gamepack2026-07-07-12-47-48-315.txt

feat: net10.0 Linux Console mode — zero Mono, 0 compile errors

2026-07-07 12:47:59 +08:00

C:\Users\Administrator\Desktop\temp\test\gamepack2026-07-07-12-48-41-060.txt

fix: Windows paths → ./gamepack/, empty Console.ReadLine guard

2026-07-07 12:56:39 +08:00

C:\Users\Administrator\Desktop\temp\test\gamepack2026-07-07-12-49-13-991.txt

fix: Windows paths → ./gamepack/, empty Console.ReadLine guard

2026-07-07 12:56:39 +08:00

Directory.Packages.props

feat: NeuralBot — ONNX模型驱动的Bot + 混合Bot训练架构

2026-07-13 02:20:44 +08:00

global.json

fix: global.json add rollForward latestMajor for Mac SDK 10.0.301

2026-07-13 03:41:20 +08:00

model.onnx

fix: 训练编排器 — V训练从头开始 + ONNX外部数据文件复制

2026-07-13 08:14:05 +08:00

model.onnx.data

fix: 训练编排器 — V训练从头开始 + ONNX外部数据文件复制

2026-07-13 08:14:05 +08:00

SeversCode.sln

feat: initial commit - HJHA game server full source

2026-07-07 12:02:15 +08:00

Description

C# HJHA Card Game Server — 生产级棋牌游戏服务器框架，支持25+玩法

35 MiB

Languages

C# 99.6%

Python 0.4%