Upload 15 files

dataset1000/README.md

@@ -0,0 +1,96 @@
+# Go Games Dataset for PyTorch Neural Network Training
+
+## Overview
+
+This dataset contains Go game positions extracted from high-quality SGF files for training neural networks. The positions are organized into three strength categories based on game quality.
+
+## Dataset Statistics
+
+- **Total SGF Files Processed**: 61149
+- **Valid SGF Files**: 0
+- **Total Positions**: 29884
+- **Processing Time**: 14.90 seconds
+
+## Strength Categories
+
+The dataset is divided into three strength categories:
+
+- **Standard** (Quality 80-85): 2704 games, 9934 positions
+- **Strong** (Quality 86-92): 3397 games, 9958 positions
+- **Elite** (Quality 93-100): 55048 games, 9992 positions
+
+## Directory Structure
+
+```
+dataset/
+├── train/
+│   ├── boards.pt      # Board state tensors (N, C, H, W)
+│   ├── moves.pt       # Move labels (N,)
+│   ├── colors.pt      # Player colors (N,)
+│   └── metadata.json  # Additional information
+├── val/
+│   ├── boards.pt
+│   ├── moves.pt
+│   ├── colors.pt
+│   └── metadata.json
+├── test/
+│   ├── boards.pt
+│   ├── moves.pt
+│   ├── colors.pt
+│   └── metadata.json
+├── stats.json         # Processing statistics
+└── README.md          # This file
+```
+
+## Board Representation
+
+The board state is represented as a tensor with 3 channels:
+1. Black stones (1 where black stone is present, 0 elsewhere)
+2. White stones (1 where white stone is present, 0 elsewhere)
+3. Next player (all 1s if black to play, all 0s if white to play)
+
+## Usage with PyTorch
+
+```python
+import torch
+import json
+import os
+from torch.utils.data import Dataset, DataLoader
+
+class GoDataset(Dataset):
+    def __init__(self, data_dir):
+        self.boards = torch.load(os.path.join(data_dir, "boards.pt"))
+        self.moves = torch.load(os.path.join(data_dir, "moves.pt"))
+        self.colors = torch.load(os.path.join(data_dir, "colors.pt"))
+        
+        with open(os.path.join(data_dir, "metadata.json"), 'r', encoding='utf-8') as f:
+            self.metadata = json.load(f)
+    
+    def __len__(self):
+        return len(self.moves)
+    
+    def __getitem__(self, idx):
+        return {
+            'board': self.boards[idx],
+            'move': self.moves[idx],
+            'color': self.colors[idx]
+        }
+
+# Create datasets
+train_dataset = GoDataset('dataset/train')
+val_dataset = GoDataset('dataset/val')
+test_dataset = GoDataset('dataset/test')
+
+# Create data loaders
+train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
+val_loader = DataLoader(val_dataset, batch_size=64)
+test_loader = DataLoader(test_dataset, batch_size=64)
+```
+
+## License
+
+The dataset is intended for research and educational purposes only.
+
+## Creation Date
+
+This dataset was created on 2025.3.13

dataset1000/README_CN.md

@@ -0,0 +1,151 @@
+# 围棋对弈数据集（PyTorch神经网络训练专用）
+
+## 概述
+
+本数据集包含从高质量SGF棋谱文件中提取的围棋对局位置，专为训练神经网络而设计。位置根据对局质量分为三个强度类别，每个类别提取了约1000个样本。
+
+## 数据集统计
+
+- **处理的SGF文件总数**：根据原始文件数量而定
+- **有效SGF文件**：通过质量筛选的文件数
+- **总位置数**：大约3000个（每个强度类别约1000个）
+- **处理时间**：取决于实际运行耗时
+
+## 强度类别
+
+数据集根据棋谱质量分为三个强度类别：
+
+- **标准级别** (Quality 80-85)：业余高段和职业初段对局
+- **强力级别** (Quality 86-92)：职业中高段对局
+- **精英级别** (Quality 93-100)：顶尖职业选手对局
+
+## 目录结构
+
+```
+dataset/
+├── train/
+│   ├── boards.pt      # 棋盘状态张量 (N, C, H, W)
+│   ├── moves.pt       # 着法标签 (N,)
+│   ├── colors.pt      # 棋手颜色 (N,)
+│   └── metadata.json  # 附加信息
+├── val/
+│   ├── boards.pt
+│   ├── moves.pt
+│   ├── colors.pt
+│   └── metadata.json
+├── test/
+│   ├── boards.pt
+│   ├── moves.pt
+│   ├── colors.pt
+│   └── metadata.json
+├── stats.json         # 处理统计信息
+└── README.md          # 本文件
+```
+
+## 棋盘表示
+
+棋盘状态表示为具有3个通道的张量：
+1. 黑棋（黑子位置为1，其他位置为0）
+2. 白棋（白子位置为1，其他位置为0）
+3. 下一手（黑方行棋时全部为1，白方行棋时全部为0）
+
+## PyTorch使用示例
+
+```python
+import torch
+import json
+import os
+from torch.utils.data import Dataset, DataLoader
+
+class GoDataset(Dataset):
+    def __init__(self, data_dir):
+        self.boards = torch.load(os.path.join(data_dir, "boards.pt"))
+        self.moves = torch.load(os.path.join(data_dir, "moves.pt"))
+        self.colors = torch.load(os.path.join(data_dir, "colors.pt"))
+        
+        with open(os.path.join(data_dir, "metadata.json"), 'r', encoding='utf-8') as f:
+            self.metadata = json.load(f)
+    
+    def __len__(self):
+        return len(self.moves)
+    
+    def __getitem__(self, idx):
+        return {
+            'board': self.boards[idx],
+            'move': self.moves[idx],
+            'color': self.colors[idx]
+        }
+
+# 创建数据集
+train_dataset = GoDataset('dataset/train')
+val_dataset = GoDataset('dataset/val')
+test_dataset = GoDataset('dataset/test')
+
+# 创建数据加载器
+train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
+val_loader = DataLoader(val_dataset, batch_size=64)
+test_loader = DataLoader(test_dataset, batch_size=64)
+```
+
+## 模型训练示例
+
+以下是使用该数据集训练简单围棋策略网络的示例代码：
+
+```python
+import torch
+import torch.nn as nn
+import torch.optim as optim
+
+# 定义一个简单的围棋策略网络
+class SimplePolicyNet(nn.Module):
+    def __init__(self, board_size=19):
+        super(SimplePolicyNet, self).__init__()
+        self.conv1 = nn.Conv2d(3, 64, 3, padding=1)
+        self.conv2 = nn.Conv2d(64, 64, 3, padding=1)
+        self.conv3 = nn.Conv2d(64, 128, 3, padding=1)
+        self.fc = nn.Linear(128 * board_size * board_size, board_size * board_size)
+        
+    def forward(self, x):
+        x = torch.relu(self.conv1(x))
+        x = torch.relu(self.conv2(x))
+        x = torch.relu(self.conv3(x))
+        x = x.view(x.size(0), -1)
+        x = self.fc(x)
+        return x
+
+# 初始化模型、损失函数和优化器
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = SimplePolicyNet().to(device)
+criterion = nn.CrossEntropyLoss()
+optimizer = optim.Adam(model.parameters(), lr=0.001)
+
+# 训练循环
+for epoch in range(10):
+    model.train()
+    running_loss = 0.0
+    
+    for batch in train_loader:
+        boards = batch['board'].to(device)
+        moves = batch['move'].to(device)
+        
+        optimizer.zero_grad()
+        outputs = model(boards)
+        loss = criterion(outputs, moves)
+        loss.backward()
+        optimizer.step()
+        
+        running_loss += loss.item()
+    
+    print(f"Epoch {epoch+1}, Loss: {running_loss/len(train_loader):.4f}")
+
+# 保存模型
+torch.save(model.state_dict(), "go_policy_model.pth")
+```
+
+## 使用许可
+
+本数据集仅供研究和教育目的使用。
+
+## 创建日期
+
+数据集创建于：2025-03-13

dataset1000/stats.json

@@ -0,0 +1,17 @@
+{
+  "total_sgf_files": 61149,
+  "valid_sgf_files": 0,
+  "processed_games": 0,
+  "total_positions": 29884,
+  "positions_by_category": {
+    "standard": 9934,
+    "strong": 9958,
+    "elite": 9992
+  },
+  "games_by_category": {
+    "standard": 2704,
+    "strong": 3397,
+    "elite": 55048
+  },
+  "processing_time": 14.901111
+}

dataset1000/test/boards.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aed553001b0386325a0f7165a08928ba6725f7f512fa354eb2d220529c948019
+size 12953815

dataset1000/test/colors.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58ef6d6ed61d9e73521cb401c564e35a935a95be93de8e536834afed5a17df6d
+size 4119

dataset1000/test/moves.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9517910984f011eba6123adddec4ade022caa1ed2b3859c7aa358aedc0338d2
+size 25042

dataset1000/train/boards.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea16d6e9336f3c2b6bb9cae26bd28e7feaa9a9c9f76d2eb4e84afc3f2307ba43
+size 103561943

dataset1000/train/colors.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:223d51806905b9fc6e2b1a1196f6d2f48cc984a881f8e54eb24044044d3749ce
+size 25047

dataset1000/train/moves.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a505415d7bc46d38b59b136abbdb50ca915ae25384fdef30a0b7c87eaa6c6ff
+size 192402

dataset1000/val/boards.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea3cbb5460898c8c8f0acf42ba5fe35406e22f4a73ac182a595b8e872124d276
+size 12945175

dataset1000/val/colors.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b185a613783b5afeea2e72c2411416eef0780f8778c694278772adceeee418d
+size 4119

dataset1000/val/moves.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbd036aadac6afae9b97e7abb84295c31e607092cedf935c99b58c371e16b0ea
+size 25042