Runtime Error

[zhixwang]

Runtime Error

[InfamousMeGa]

Hi zhixwang,

Sorry that we have never met this issue before. According to the error information, seems like the CUDA libs on your Windows machine are either not correctly installed or not compatible with neural-ilt.

My general suggestion is to prepare and test Neural-ilt on a linux machine with CUDA correctly installed. FYI, the code in this repo can be run on our side with the following evn setups:
Ubuntu 18.04.6 LTS (GNU/Linux 4.15.0-158-generic x86_64)
CUDA Version: 11.4

Thanks!

[zhixwang]

Hi authors, could you please also share the version of cv2(opencv) library for the repo?

I tried to run python neural_ilt.py and got the following error info:

Namespace(beta=1.45, gpu_no=0, load_model_name='iccad_32nm_m1_wts.pth', select_by_obj=True)
Launching Neural-ILT on device: cuda:0
-------- Loading Neural-ILT Model & Data --------
MODEL: models/unet/iccad_32nm_m1_wts.pth
DATASET: ICCAD2013-IBM-Benchmark

Processing t1_0_mask.png with size of (245, 291, 1537, 1583) and scale factor = [2.5234375, 2.5234375]

--- Initializing Model for t1_0_mask.png ---
Traceback (most recent call last):
  File "neural_ilt.py", line 436, in <module>
    run_neural_ilt_ibm_bench()
  File "neural_ilt.py", line 356, in run_neural_ilt_ibm_bench
    l2_avg, pv_avg, epe_avg, runtime_avg = nerual_ilt.neural_ilt_correction(refine_data_loader)
  File "neural_ilt.py", line 185, in neural_ilt_correction
    inputs, labels, new_cord
  File "/root/anaconda3/envs/nilt/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/root/neural-ilt/neural_ilt_backbone.py", line 104, in forward
    out_loss = self.ilt_loss_layer(x, y, new_cord)
  File "/root/anaconda3/envs/nilt/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/root/neural-ilt/ilt_loss_layer.py", line 164, in forward
    return ilt_loss_scale_function.apply(preds, target, self.kernels, self.kernels_ct, self.kernel_def, self.kernel_def_ct, self.weight, self.weight_def,new_cord, self.cycle_mode, self.cplx_obj, self.report_epe)
  File "/root/neural-ilt/ilt_loss_layer.py", line 116, in forward
    checkpoints = get_epe_checkpoints((target.detach().data.cpu().numpy()[0][0] * 255).astype(np.uint8))
  File "/root/neural-ilt/utils/epe_checker.py", line 64, in get_epe_checkpoints
    polys, _ = find_all_contours(layout)
  File "/root/neural-ilt/utils/epe_checker.py", line 16, in find_all_contours
    cnts, hier = cv.findContours(gray_img, cv.RETR_TREE, contour_approx)
ValueError: too many values to unpack (expected 2)

I am running the code on a Linux machine with GPU. All the packages are installed with the correct version as required by the repo, except for opencv which is the default version (3.4.2). I suspect that this error might be caused by a wrong version of opencv library.

Thanks!

[zhixwang]

I managed to run the repo with opencv=4.3.0.

Thanks!