import sys
from collections import OrderedDict
sys.path.append(r"./RepVGG")
import torch
import torch.onnx
from repvgg import get_RepVGG_func_by_name
def load_my_state_dict(model, state_dict):
own_state = model.state_dict()
if 'state_dict' in state_dict:
state_dict = state_dict['state_dict']
for name, param in state_dict.items():
if name not in own_state:
if name.startswith("module."):
own_state[name.split("module.")[-1]].copy_(param)
else:
print(name, " not loaded")
continue
else:
own_state[name].copy_(param)
return model
def convert(checkpoint, output_file):
repvgg_build_func = get_RepVGG_func_by_name("RepVGG-A0")
model = repvgg_build_func(deploy=False)
model = load_my_state_dict(
model, torch.load(checkpoint, map_location='cpu'))
model.eval()
input_names = ["actual_input_1"]
output_names = ["output1"]
dynamic_axes = {"actual_input_1": {0: "-1"}, "output1": {0: "-1"}}
dummy_input = torch.randn(1, 3, 224, 224)
torch.onnx.export(
model,
dummy_input,
output_file,
input_names=input_names,
output_names=output_names,
dynamic_axes=dynamic_axes,
opset_version=11
)
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser('data preprocess.')
parser.add_argument('--checkpoint', type=str, default=None,
help='path to PyTorch pretrained file(.pth)')
parser.add_argument('--onnx', type=str, default='./RepVGG.onnx',
help='path to save onnx model(.onnx)')
args = parser.parse_args()
convert(args.checkpoint, args.onnx)