Floating point exception (core dumped) for UNet implementation
Question:
I am trying to do an implementation of KiuNet ( https://github.com/jeya-maria-jose/KiU-Net-pytorch ). But when I am executing the train command like so:
python train.py --train_dataset "KiuNet/Train Folder/" --val_dataset "KiuNet/Validation Folder/" --direc 'KiuNet/Results/' --batch_size 1 --epoch 200 --save_freq 10 --modelname "kiunet" --learning_rate 0.0001
I am getting the following error:
Traceback (most recent call last):
File "KiuNet/KiU-Net-pytorch/train.py", line 235, in <module>
loss.backward()
File "/miniconda3/lib/python3.9/site-packages/torch/_tensor.py", line 487, in backward
torch.autograd.backward(
File "/miniconda3/lib/python3.9/site-packages/torch/autograd/__init__.py", line 197, in backward
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [847,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [958,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [703,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [830,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [831,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [575,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [974,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [77,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [78,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [719,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [720,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [592,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [593,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [209,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [465,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [337,0,0] Assertion `t >= 0 && t < n_classes` failed.
When I am running the train command with CUDA_LAUNCH_BLOCKING=1 I get the following error:
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [840,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [580,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [453,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [326,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [71,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [712,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [198,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [199,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [968,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [959,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [830,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [574,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [702,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [191,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [318,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [319,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [446,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [63,0,0] Assertion `t >= 0 && t < n_classes` failed.
Floating point exception (core dumped)
My torch and CUDA version are: '1.13.0+cu117'
My Python version: Python 3.9.12
Any help is much appreciated!
Answers:
The repository author mentions the following.
"This bug occurs when the ground truth masks have more classes than the number of classes in prediction. Please make sure you ground truth images have only 0 or 1 labels of pixels if you are training for binary segmentation. The datasets usually have the ground truth as 0 or 255 labels of pixels. So, please convert them to 0’s and 1’s."
I am trying to do an implementation of KiuNet ( https://github.com/jeya-maria-jose/KiU-Net-pytorch ). But when I am executing the train command like so:
python train.py --train_dataset "KiuNet/Train Folder/" --val_dataset "KiuNet/Validation Folder/" --direc 'KiuNet/Results/' --batch_size 1 --epoch 200 --save_freq 10 --modelname "kiunet" --learning_rate 0.0001
I am getting the following error:
Traceback (most recent call last):
File "KiuNet/KiU-Net-pytorch/train.py", line 235, in <module>
loss.backward()
File "/miniconda3/lib/python3.9/site-packages/torch/_tensor.py", line 487, in backward
torch.autograd.backward(
File "/miniconda3/lib/python3.9/site-packages/torch/autograd/__init__.py", line 197, in backward
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [847,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [958,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [703,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [830,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [831,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [575,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [974,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [77,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [78,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [719,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [720,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [592,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [593,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [209,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [465,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [337,0,0] Assertion `t >= 0 && t < n_classes` failed.
When I am running the train command with CUDA_LAUNCH_BLOCKING=1 I get the following error:
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [840,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [580,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [453,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [326,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [71,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [712,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [198,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [199,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [968,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [959,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [830,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [574,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [702,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [191,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [318,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [319,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [446,0,0] Assertion `t >= 0 && t < n_classes` failed.
../aten/src/ATen/native/cuda/NLLLoss2d.cu:104: nll_loss2d_forward_kernel: block: [0,0,0], thread: [63,0,0] Assertion `t >= 0 && t < n_classes` failed.
Floating point exception (core dumped)
My torch and CUDA version are: '1.13.0+cu117'
My Python version: Python 3.9.12
Any help is much appreciated!
The repository author mentions the following.
"This bug occurs when the ground truth masks have more classes than the number of classes in prediction. Please make sure you ground truth images have only 0 or 1 labels of pixels if you are training for binary segmentation. The datasets usually have the ground truth as 0 or 255 labels of pixels. So, please convert them to 0’s and 1’s."