bench.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from functorch import vmap, combine_state_for_ensemble
from torch.utils.benchmark import Timer

class LeNet5(nn.Module):
    def __init__(self):
        super(LeNet5, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1 = nn.Linear(400, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 400)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

device = 'cuda'
models = [LeNet5().to(device) for _ in range(5)]
fmodel, params, buffers = combine_state_for_ensemble(models)

# cifar10 dataset: batch size 128, 3 channel, 32x32 images
data = torch.randn(128, 3, 32, 32).to(device)

def vmap_inference():
    results = vmap(fmodel, (0, 0, None))(params, buffers, data)
    return results

def forloop_inference():
    results = []
    for model in models:
        results.append(model(data))
    return torch.stack(results)

t0 = Timer('vmap_inference()', setup='from __main__ import vmap_inference')
t1 = Timer('forloop_inference()', setup='from __main__ import forloop_inference')

print(t0.timeit(1000))
print(t1.timeit(1000))