error when export to a onnx model

[bigsheep2012]

Hello CSAILVision team,
Thanks for sharing this work.

I trained a model with 'resnet18dilated' and 'ppm_deepsup'.
Trying to export this model to a onnx model and get errors (I can export a normal ResNet50 to a onnx model normally).

The PSPNet contains your source files: lib folder, models.py and resnet.py
I can train the net successfully.

Please check the code for producing the error:
from PSPNet import ModelBuilder, SegmentationModule
builder = ModelBuilder()
net_encoder = builder.build_encoder(arch='resnet18dilated', fc_dim=512, weights='')
net_decoder = builder.build_decoder(arch='ppm_deepsup',fc_dim=512, num_class=150, weights='', use_softmax=False)
crit = nn.NLLLoss(ignore_index=-1)
model = SegmentationModule(net_encoder, net_decoder, crit)

dummy_input = torch.randn(1, 3, 640, 640, device='cuda')
model = model.cuda()

state_dict = torch.load('./weights/PSP_scratch.pth')
model.load_state_dict(state_dict)

input_names = [ "input" ]
output_names = [ "output" ]
torch.onnx.export(model, dummy_input, "model.onnx", verbose=True, input_names=input_names, output_names=output_names)

Many thanks

the errors:
File "transform_to_onnx.py", line 53, in
torch.onnx.export(model, dummy_input, "model.onnx", verbose=True, input_names=input_names, output_names=output_names)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/init.py", line 27, in export
return utils.export(*args, **kwargs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 104, in export
operator_export_type=operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 281, in _export
example_outputs, propagate)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 227, in _model_to_graph
graph = _optimize_graph(graph, operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 155, in _optimize_graph
graph = torch._C._jit_pass_onnx(graph, operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/init.py", line 52, in _run_symbolic_function
return utils._run_symbolic_function(*args, **kwargs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 504, in _run_symbolic_function
return fn(g, *inputs, **attrs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/symbolic.py", line 88, in wrapper
args = [_parse_arg(arg, arg_desc) for arg, arg_desc in zip(args, arg_descriptors)]
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/symbolic.py", line 45, in _parse_arg
raise RuntimeError("ONNX symbolic expected a constant value in the trace")
RuntimeError: ONNX symbolic expected a constant value in the trace
(py27) yanglin@yanglin-pc:/media/yanglin/Lindisk/Robosense/git_projects/mjj2/sti_predict/src/cnn_detection/python/pixel_based/apollo_fcn_box$ python transform_to_onnx.py
Traceback (most recent call last):
File "transform_to_onnx.py", line 53, in
torch.onnx.export(model, dummy_input, "model.onnx", verbose=True, input_names=input_names, output_names=output_names)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/init.py", line 27, in export
return utils.export(*args, **kwargs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 104, in export
operator_export_type=operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 281, in _export
example_outputs, propagate)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 227, in _model_to_graph
graph = _optimize_graph(graph, operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 155, in _optimize_graph
graph = torch._C._jit_pass_onnx(graph, operator_export_type)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/init.py", line 52, in _run_symbolic_function
return utils._run_symbolic_function(*args, **kwargs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/utils.py", line 504, in _run_symbolic_function
return fn(g, *inputs, **attrs)
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/symbolic.py", line 88, in wrapper
args = [_parse_arg(arg, arg_desc) for arg, arg_desc in zip(args, arg_descriptors)]
File "/home/yanglin/anaconda3/envs/py27/lib/python2.7/site-packages/torch/onnx/symbolic.py", line 45, in _parse_arg
raise RuntimeError("ONNX symbolic expected a constant value in the trace")
RuntimeError: ONNX symbolic expected a constant value in the trace

[wm901115nwpu]

how to export decoder onnx

[wm901115nwpu]

RuntimeError: Expected 4-dimensional input for 4-dimensional weight [128, 512, 3, 3], but got 3-dimensional input of size [512, 63, 63] instead

[Jason-xin]

Have you solved this problems? And there are two models including encoder and decoder, how can you connect this two model when exporting?

[ShihuaiXu]

anyone solve it? please help my qq is 531232693

[ynma-hanvo]

also want to export it to onnx ,but got no idea.
any help?

[Md-Sayeed-Khan]

how to convert to onnx ? getting issue ONNX export of operator adaptive_avg_pool2d, output size that are not factor of input size.
please help