def test_model_inference():
# get all models
from resnest.gluon.model_store import _model_sha1
from resnest.gluon import get_model
model_list = _model_sha1.keys()
x = mx.random.uniform(shape=(1, 3, 224, 224))
for model_name in model_list:
print('Doing: ', model_name)
model = get_model(model_name, pretrained=True)
y = model(x)
num_gpus = opt.num_gpus
if num_gpus > 0:
batch_size *= num_gpus
ctx = [mx.gpu(i) for i in range(num_gpus)] if num_gpus > 0 else [mx.cpu()]
num_workers = opt.num_workers
input_size = opt.crop_size
model_name = opt.model
pretrained = True if not opt.params_file else False
kwargs = {'ctx': ctx, 'pretrained': pretrained, 'classes': classes}
if opt.dilation > 1:
kwargs['dilation'] = opt.dilation
net = get_model(model_name, **kwargs)
net.cast(opt.dtype)
if opt.params_file:
net.load_parameters(opt.params_file, ctx=ctx)
else:
net.hybridize()
normalize = transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
crop_ratio = opt.crop_ratio if opt.crop_ratio > 0 else 0.875
resize = int(math.ceil(input_size / crop_ratio))
if input_size > 320:
transform_test = transforms.Compose([
ToPIL(),
ECenterCrop(input_size),
# Get model from GluonCV model zoo
# https://gluon-cv.mxnet.io/model_zoo/index.html
kwargs = {
'ctx': context,
'pretrained': args.use_pretrained,
'classes': num_classes,
'input_size': args.input_size
}
if args.last_gamma:
kwargs['last_gamma'] = True
if args.dropblock_prob > 0:
kwargs['dropblock_prob'] = args.dropblock_prob
net = get_model(args.model, **kwargs)
net.cast(args.dtype)
from gluoncv.nn.dropblock import DropBlockScheduler
# does not impact normal model
drop_scheduler = DropBlockScheduler(net, 0, 0.1, args.num_epochs)
if rank == 0:
logging.info(net)
# Create initializer
initializer = mx.init.Xavier(rnd_type='gaussian',
factor_type="in",
magnitude=2)