Consumes 25% of RAM

[sathisha-beltech]

When I executed the following usage code given in the README file, I found that the code is consuming 25% of RAM.

import torch
from torch import nn
from torchvision.models.resnet import resnet50

# create some regular pytorch model...
model = resnet50().cuda().eval()

# create example data
x = torch.ones((1, 3, 224, 224)).cuda()

# convert to TensorRT feeding sample data as input
opt_shape_param = [
    [
        [1, 3, 128, 128],   # min
        [1, 3, 256, 256],   # opt
        [1, 3, 512, 512]    # max
    ]
]
model_trt = torch2trt_dynamic(model, [x], fp16_mode=False, opt_shape_param=opt_shape_param)