GaryGuTC/COMG_model

core dumped

ecust-sl opened this issue · 5 comments

ecust-sl commented

train_iu_xray.sh: line 21: 1749424 Segmentation fault (core dumped) python main_train.py --image_dir /extra/shilei/dataset/physionet.org/files/iu_xray/images --ann_path /extra/shilei/dataset/physionet.org/files/iu_xray/annotation_disease.json --dataset_name iu_xray --max_seq_length 60 --threshold 3 --epochs 150 --batch_size 8 --lr_ve 1e-4 --lr_ed 5e-4 --step_size 10 --gamma 0.8 --num_layers 3 --topk 32 --cmm_size 2048 --cmm_dim 512 --seed 7580 --beam_size 3 --save_dir results/iu_xray/ --log_period 50 --early_stop 40
Why is this a problem?

GaryGuTC commented

Hi

I haven't seen this error before. Could you provide some extra details (like screenshots without any sensitive information)?

Best regards,
Tiancheng

ecust-sl commented

Hi

I haven't seen this error before. Could you provide some extra details (like screenshots without any sensitive information)?

Best regards, Tiancheng

image 这个是报错的图片，可以帮忙看一下问题貌似出现在resnet101那里
ecust-sl commented

Hi
I haven't seen this error before. Could you provide some extra details (like screenshots without any sensitive information)?
Best regards, Tiancheng

image 这个是报错的图片，可以帮忙看一下问题貌似出现在resnet101那里

image 在这里调用resnet101的时候报core dumped
GaryGuTC commented

I suggest that

  1. Check the pytorch version. (I used torch 2.0.1 + cuda==11.8)
  2. Please refer this issue.

Hope this can help you
Best regards

ecust-sl commented

I suggest that

  1. Check the pytorch version. (I used torch 2.0.1 + cuda==11.8)
  2. Please refer this issue.

Hope this can help you Best regards

Ok, thank you, I'm using torch=2.2.1 + cuda11.8 I'll try to modify the version of the torch