安装指南¶

#!/usr/bin/env python3
"""Genesis安装验证脚本"""

def test_basic_import():
    """测试基础导入"""
    try:
        import genesis
        import genesis.nn as nn
        import genesis.optim as optim
        print("✅ Genesis导入成功")
        print(f"   核心模块: genesis, nn, optim")
        print(f"   可用函数: {len([x for x in dir(genesis) if not x.startswith('_')])}")
    except ImportError as e:
        print(f"❌ Genesis导入失败: {e}")
        return False
    return True

def test_tensor_operations():
    """测试张量操作"""
    try:
        import genesis

        # 创建张量
        x = genesis.randn(3, 4)
        y = genesis.randn(3, 4)

        # 基础运算
        z = x + y
        w = genesis.matmul(x, y.T)  # 使用实际的Genesis API

        print("✅ 张量运算正常")
        print(f"   加法结果形状: {z.shape}")
        print(f"   矩阵乘法形状: {w.shape}")
    except Exception as e:
        print(f"❌ 张量运算失败: {e}")
        return False
    return True

def test_neural_networks():
    """测试神经网络模块"""
    try:
        import genesis
        import genesis.nn as nn

        # 使用实际的Genesis模块创建简单模型
        model = nn.Sequential([
            nn.Linear(10, 5),
            nn.ReLU(),
            nn.Linear(5, 1)
        ])

        # 测试前向传播
        x = genesis.randn(2, 10)
        y = model(x)
        print("✅ 神经网络模块正常")
        print(f"   模型层数: {len(list(model.parameters()))} 个参数张量")
        print(f"   输出形状: {y.shape}")
    except Exception as e:
        print(f"❌ 神经网络模块失败: {e}")
        return False
    return True

def test_backend_support():
    """测试后端支持"""
    try:
        import genesis
        from genesis.backend import default_device

        # 测试基础后端功能
        device = default_device()
        x = genesis.randn(5, 5)

        print("✅ 后端支持正常")
        print(f"   默认设备: {device}")
        print(f"   张量设备: {x.device}")

        # 尝试检测CUDA是否可用
        try:
            # 测试是否可以创建CUDA张量
            import torch
            if torch.cuda.is_available():
                print("   检测到CUDA（通过PyTorch后端）")
            else:
                print("   CUDA不可用（仅CPU）")
        except:
            print("   后端: Genesis原生")

    except Exception as e:
        print(f"❌ 后端测试失败: {e}")
        return False
    return True

def test_autograd():
    """测试自动微分"""
    try:
        import genesis

        # 测试基础自动微分
        x = genesis.randn(5, requires_grad=True)
        y = genesis.functional.sum(x * x)  # 使用实际的Genesis API
        y.backward()

        print("✅ 自动微分正常")
        print(f"   输入形状: {x.shape}")
        print(f"   梯度已计算: {x.grad is not None}")
        print(f"   梯度形状: {x.grad.shape if x.grad is not None else 'None'}")
    except Exception as e:
        print(f"❌ 自动微分失败: {e}")
        return False
    return True

def test_optimizers():
    """测试优化器功能"""
    try:
        import genesis
        import genesis.nn as nn
        import genesis.optim as optim

        # 创建简单模型和优化器
        model = nn.Linear(5, 1)
        optimizer = optim.Adam(model.parameters(), lr=0.01)

        # 测试基础优化步骤
        x = genesis.randn(3, 5)
        y_pred = model(x)
        loss = genesis.functional.sum(y_pred * y_pred)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        print("✅ 优化器功能正常")
        print(f"   优化器类型: {type(optimizer).__name__}")
        print(f"   学习率: 0.01")
        print(f"   参数已更新: {len(list(model.parameters()))}")
    except Exception as e:
        print(f"❌ 优化器测试失败: {e}")
        return False
    return True

def test_serialization():
    """测试模型保存/加载"""
    try:
        import genesis
        import genesis.nn as nn

        # 创建并保存模型
        model = nn.Linear(3, 2)
        state_dict = model.state_dict()

        # 测试序列化功能
        genesis.save(state_dict, 'test_model.pkl')
        loaded_state = genesis.load('test_model.pkl')

        print("✅ 序列化功能正常")
        print(f"   模型保存和加载成功")
        print(f"   状态字典键数量: {len(state_dict)}")

        # 清理
        import os
        if os.path.exists('test_model.pkl'):
            os.remove('test_model.pkl')

    except Exception as e:
        print(f"❌ 序列化测试失败: {e}")
        return False
    return True

if __name__ == "__main__":
    print("🔍 Genesis安装验证\n")

    tests = [
        test_basic_import,
        test_tensor_operations,
        test_neural_networks,
        test_backend_support,
        test_autograd,
        test_optimizers,
        test_serialization
    ]

    passed = 0
    total = len(tests)

    for test in tests:
        try:
            if test():
                passed += 1
        except Exception as e:
            print(f"❌ 测试异常失败: {e}")
        print()

    print(f"📊 测试结果: {passed}/{total} 通过")

    if passed == total:
        print("🎉 恭喜！Genesis安装成功，所有功能正常！")
    elif passed >= total * 0.8:  # 80%通过率
        print("✅ Genesis安装基本成功！检测到少量问题。")
        print("   大部分功能正常运行。请检查上面的失败测试。")
    else:
        print("⚠️  Genesis安装存在问题。请检查：")
        print("   1. Genesis正确安装: pip install -e .")
        print("   2. 依赖已安装: pip install torch triton")
        print("   3. Python版本为3.8+")

import genesis

# 启用内存优化
genesis.cuda.empty_cache()

# 减小批量大小
batch_size = 16  # 替代原来的32

# 启用梯度检查点 (如果支持)
model.gradient_checkpointing = True