Device & Platform Utilities#

用于检测和配置计算平台和设备的函数。

设备信息#

import genesis as gs

gs.init(backend=gs.gpu)

# Get PyTorch device
device = gs.device
print(device)  # cuda:0, mps:0, cpu, etc.

# Get active backend
backend = gs.backend
print(backend)  # gs.cuda, gs.metal, gs.cpu, etc.

Backend 选择#

自动选择#

# GPU auto-selects based on platform
gs.init(backend=gs.gpu)
# Linux -> CUDA
# macOS -> Metal

手动选择#

# Force specific backend
gs.init(backend=gs.cuda)    # NVIDIA CUDA
gs.init(backend=gs.metal)   # Apple Metal
gs.init(backend=gs.cpu)     # CPU fallback

随机种子#

import genesis as gs

# Set random seed for reproducibility
gs.init(seed=42)

# Or set later
gs.set_random_seed(42)

全局变量#

gs.init() 后，以下变量可用：

Variable

Type

Description

gs.device

torch.device

PyTorch 张量设备

gs.backend

gs.backend

活动计算后端

gs.EPS

float

当前精度下的机器 epsilon

类型提示#

# Quadrants types
gs.qd_float  # Quadrants float type
gs.qd_int    # Quadrants int type
gs.qd_vec3   # Quadrants 3D vector
gs.qd_mat3   # Quadrants 3x3 matrix

# PyTorch types
gs.tc_float  # PyTorch float dtype
gs.tc_int    # PyTorch int dtype

另请参阅#