search

Amshaker / unetr_plus_plus

[IEEE TMI-2024] UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation
Apache License 2.0
360 stars 34 forks source link

Task003_tumor dataset cannot be trained? #55

Open liaochuanlin opened 1 year ago

liaochuanlin commented 1 year ago

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 commented 10 months ago

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

Amshaker commented 9 months ago

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?