I am trying out the following vgg11 example from torch hub. I have modified it using psutil to capture process memory usage at different stages.
Observtion - Post inference/prediction call, the model memory increases e.g. for vgg11 ->
pre inference/after load call -> ~691.62890625MB
post inference/after forward call -> ~747.203125MB
Question -> Is this an expected behavior and why? Or am I missing something in following code snippet?
import torch
import sys
import os
import psutil
import gc
def memory_usage_psutil():
# return the memory usage in percentage like top
process = psutil.Process(os.getpid())
mem = process.memory_info()[0] / float(2** 20)
#mem = process.memory_percent()
return mem
from time import sleep
mem1 = str(memory_usage_psutil())
model = torch.hub.load('pytorch/vision:v0.6.0', 'vgg11', pretrained=True)
model.eval()
mem2 = str(memory_usage_psutil())
import urllib
url, filename = ("https://github.com/pytorch/hub/raw/master/dog.jpg", "dog.jpg")
try: urllib.URLopener().retrieve(url, filename)
except: urllib.request.urlretrieve(url, filename)
from PIL import Image
from torchvision import transforms
input_image = Image.open(filename)
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
input_tensor = preprocess(input_image)
input_batch = input_tensor.unsqueeze(0) # create a mini-batch as expected by the model
mem3 = str(memory_usage_psutil())
# move the input and model to GPU for speed if available
if torch.cuda.is_available():
input_batch = input_batch.to('cuda')
model.to('cuda')
with torch.no_grad():
model(input_batch)
mem4 = str(memory_usage_psutil())
gc.collect()
sleep(60)
mem5 = str(memory_usage_psutil())
print(mem1+','+mem2+','+mem3+','+mem4+','+mem5)
# Tensor of shape 1000, with confidence scores over Imagenet's 1000 classes
#print(output[0])
# The output has unnormalized scores. To get probabilities, you can run a softmax on it.
#print(torch.nn.functional.softmax(output[0], dim=0))
Thanks!