david9ml opened a new issue #18278:
URL: https://github.com/apache/incubator-mxnet/issues/18278
## Description
I tried to convert gluon model(ssd_512_resnet50_v1_custom) to onnx, but
failed !
### Error Message
Traceback (most recent call last):
File "ssd_gluon.py", line 79, in <module>
converted_model_path = onnx_mxnet.export_model(net_sym, net_params,
[input_shape], np.float32, onnx_file, verbose=True)
File
"anaconda2/envs/Python3/lib/python3.6/site-packages/mxnet/contrib/onnx/mx2onnx/export_model.py",
line 89, in export_model
raise ValueError("Input sym and params should either be files or
objects")
ValueError: Input sym and params should either be files or objects
## To Reproduce
```
import os
import mxnet as mx
import numpy as np
from mxnet import gluon, sym
from gluoncv.model_zoo import get_model
from mxnet.contrib import onnx as onnx_mxnet
from mxnet.contrib.onnx.mx2onnx._export_helper import load_module
# Load Model
model_name = 'ssd_512_resnet50_v1_custom'
resume_params = './ssd/ssd_512_resnet50_v1_custom_best.params'
pretrained = False
ctx = [mx.gpu(1)]
class_names = ['car','truck','bus','moto']
net = get_model(model_name, ctx=ctx, pretrained=False,
norm_layer=gluon.nn.BatchNorm, classes=class_names, transfer='voc')
net.load_parameters(resume_params, ctx = ctx)
net.set_nms(nms_thresh=0.45, nms_topk=200)
net.hybridize()
net_sym = net(sym.Variable('data'))
net_params = {name:param._reduce() for name, param in
net.collect_params().items()}
#net_params.update({})
# Standard Imagenet input - 3 channels, 224*224
input_shape = (1,3,512,512)
# Path of the output file
onnx_file = './ssd_gluon.onnx'
converted_model_path = onnx_mxnet.export_model(net_sym, net_params,
[input_shape], np.float32, onnx_file, verbose=True)
```
## Environment
Mxnet: 1.5.0
onnx: 1.6.0
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
