Only 4D input tensorsare supported (got 3D)

[Irennnne]

I got this problem when I tried to inference on my own dataset. Could you give me some advice? Thank you!

Traceback (most recent call last):
  File "train_net.py", line 317, in <module>
    launch(
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/detectron2/engine/launch.py", line 67, in launch
    mp.spawn(
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 240, in spawn
    return start_processes(fn, args, nprocs, join, daemon, start_method='spawn')
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 198, in start_processes
    while not context.join():
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 160, in join
    raise ProcessRaisedException(msg, error_index, failed_process.pid)
torch.multiprocessing.spawn.ProcessRaisedException: 

-- Process 1 terminated with the following error:
Traceback (most recent call last):
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 69, in _wrap
    fn(i, *args)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/detectron2/engine/launch.py", line 126, in _distributed_worker
    main_func(*args)
  File "/home/data4/jieming/new_open_voc/CAT-Seg/train_net.py", line 302, in main
    res = Trainer.test(cfg, model)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/detectron2/engine/defaults.py", line 608, in test
    results_i = inference_on_dataset(model, data_loader, evaluator)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/detectron2/evaluation/evaluator.py", line 158, in inference_on_dataset
    outputs = model(inputs)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/data4/jieming/new_open_voc/CAT-Seg/cat_seg/cat_seg_model.py", line 165, in forward
    return self.inference_sliding_window(batched_inputs)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/home/data4/jieming/new_open_voc/CAT-Seg/cat_seg/cat_seg_model.py", line 249, in inference_sliding_window
    image = rearrange(unfold(image), "(C H W) L-> L C H W", C=3, H=kernel)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/nn/modules/fold.py", line 294, in forward
    return F.unfold(input, self.kernel_size, self.dilation,
  File "/home/anaconda3/envs/kksan/lib/python3.8/site-packages/torch/nn/functional.py", line 4666, in unfold
    raise NotImplementedError("Input Error: Only 4D input Tensors are supported (got {}D)".format(input.dim()))
NotImplementedError: Input Error: Only 4D input Tensors are supported (got 3D)

[Qiming-Huang]

@Irennnne
Hi, I got the same issue, and solved it this way:

Modify the codes in cat_seg_model.py, lines 194,195 to:

# image = F.interpolate(images[0].unsqueeze(0), size=out_res, mode='bilinear', align_corners=False).squeeze()
# image = rearrange(unfold(image), "(C H W) L-> L C H W", C=3, H=kernel)

image = F.interpolate(images[0].unsqueeze(0), size=out_res, mode='bilinear', align_corners=False)
image = rearrange(unfold(image).squeeze(), "(C H W) L-> L C H W", C=3, H=kernel) 

and the code in cat_seg_model.py to:

# outputs = fold(outputs.flatten(1).T) / fold(unfold(torch.ones([1] + out_res, device=self.device)))

outputs = fold(outputs.flatten(1).T.unsqueeze(0)) / fold(unfold(torch.ones([1] + out_res, device=self.device).unsqueeze(0)))

Hope this helps.