오픈 소스 · GPU 가속

EvoX Runtime

GPU 가속 진화 연산 프레임워크

EvoX Runtime은 확장 가능한 진화 연산을 위한 분산 GPU 가속 프레임워크로, 단일 목표 및 다중 목표 최적화, 신경 진화, 진화 강화 학습을 위한 50개 이상의 알고리즘을 제공하며 PyTorch와 완벽하게 호환됩니다.

GitHub에서 보기 →

import torch
from evox.algorithms.pso_variants import PSO
from evox.problems.numerical import Ackley
from evox.workflows import StdWorkflow, EvalMonitor

torch.set_default_device("cuda")
# Define the algorithm
algorithm = PSO(pop_size=100, lb=-32 * torch.ones(10), ub=32 * torch.ones(10))
problem = Ackley()
monitor = EvalMonitor()
workflow = StdWorkflow(algorithm, problem, monitor)

workflow.init_step()
for i in range(100):
    workflow.step()

monitor.plot()

import torch
from evox.algorithms import RVEA
from evox.problems.numerical import DTLZ2
from evox.workflows import StdWorkflow, EvalMonitor

torch.set_default_device("cuda")
prob = DTLZ2(m=3)
pf = prob.pf()
algo = RVEA(pop_size=100, n_objs=3, lb=-torch.zeros(12), ub=torch.ones(12))
monitor = EvalMonitor()
workflow = StdWorkflow(algo, prob, monitor)

workflow.init_step()
for i in range(100):
    workflow.step()

import torch
import torch.nn as nn
from evox.algorithms import PSO
from evox.problems.neuroevolution.brax import BraxProblem
from evox.utils import ParamsAndVector
from evox.workflows import EvalMonitor, StdWorkflow

class SimpleMLP(nn.Module):
    def __init__(self):
        super().__init__()
        # observation space is 17-dim, action space is 6-dim.
        self.features = nn.Sequential(nn.Linear(17, 8), nn.Tanh(), nn.Linear(8, 6))

    def forward(self, x):
        x = self.features(x)
        return torch.tanh(x)

# Initialize the MLP model
torch.set_default_device('cuda')
model = SimpleMLP()
adapter = ParamsAndVector(dummy_model=model)

# Set the population size
POP_SIZE = 1024

# Get the bound of the PSO algorithm
model_params = dict(model.named_parameters())
pop_center = adapter.to_vector(model_params)
lb = torch.full_like(pop_center, -5)
ub = torch.full_like(pop_center, 5)

# Initialize the PSO
algorithm = PSO(pop_size=POP_SIZE, lb=lb, ub=ub)

# Initialize the Brax problem
problem = BraxProblem(
    policy=model,
    env_name="halfcheetah",
    max_episode_length=1000,
    num_episodes=3,
    pop_size=POP_SIZE,
)

monitor = EvalMonitor(topk=3)

workflow = StdWorkflow(
    algorithm=algorithm,
    problem=problem,
    monitor=monitor,
    opt_direction="max",
    solution_transform=adapter,
)

workflow.init_step()
for i in range(50):
    workflow.step()