Task003_tumor dataset cannot be trained?

[liaochuanlin]

I am using data from this folder: ../DATASET_Tumor/unetr_pp_raw/unetr_pp_raw_data/Task03_tumor/Task003_tumor/nnFormerData_plans_v2.1
###############################################
self.crop_size in init train: [128 128 128]
[0.57142857 0.28571429 0.14285714]
loading dataset
loading all case properties
2023-09-15 15:03:28.663895: Using splits from existing split file: ../DATASET_Tumor/unetr_pp_raw/unetr_pp_raw_data/Task03_tumor/Task003_tumor/splits_final.pkl
2023-09-15 15:03:28.664275: The split file contains 5 splits.
2023-09-15 15:03:28.664320: Desired fold for training: 0
2023-09-15 15:03:28.664411: This split has 387 training and 73 validation cases.
unpacking dataset
done
Traceback (most recent call last):
File "../unetr_pp/run/run_training.py", line 171, in
main()
File "../unetr_pp/run/run_training.py", line 142, in main
trainer.initialize(not validation_only)
File "/home/lcl/unetr_plus_plus/unetr_pp/training/network_training/unetr_pp_trainer_tumor.py", line 150, in initialize
self.initialize_network()
File "/home/lcl/unetr_plus_plus/unetr_pp/training/network_training/unetr_pp_trainer_tumor.py", line 189, in initialize_network
model_flops = flops.total()
File "/home/lcl/anaconda3/envs/unetr_pp/lib/python3.8/site-packages/fvcore/nn/jit_analysis.py", line 248, in total
stats = self._analyze()
File "/home/lcl/anaconda3/envs/unetr_pp/lib/python3.8/site-packages/fvcore/nn/jit_analysis.py", line 583, in _analyze
op_counts = self._op_handles[kind](inputs, outputs)
File "/home/lcl/anaconda3/envs/unetr_pp/lib/python3.8/site-packages/fvcore/nn/jit_handles.py", line 181, in einsum_flop_jit
assert len(inputs) == 2, len(inputs)
AssertionError: 3

[Holloworker]

Hi, I met the same problem. Did you solve it?

[Amshaker]

Could you please share the command you use for running?
It seems there is an issue in the input shape for measuring the FLOPs. Is the training working when you comment measuring the FLOPs?