search

mykcs / MedSAM_lite-0424

Apache License 2.0
0 stars 0 forks source link

MedSAM Lite size: 9791300 Finetuning with pretrained weights lite_medsam.pth - 打印了10次 #1

Open mykcs opened 2 months ago

mykcs commented 2 months ago

log

/Users/myk/anaconda3/envs/env-MedSAM-0422/bin/python /Users/myk/PyPjcts/MedSAM_lite-0424/train_mps-1.py 
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
wandb: Currently logged in as: mykcs (team-mykcs). Use `wandb login --relogin` to force relogin
wandb: Tracking run with wandb version 0.16.6
wandb: Run data is saved locally in /Users/myk/PyPjcts/MedSAM_lite-0424/wandb/run-20240424_023431-1m08ghs8
wandb: Run `wandb offline` to turn off syncing.
wandb: Syncing run run-0424-2
wandb: ⭐️ View project at https://wandb.ai/team-mykcs/MedSAM-lite
wandb: 🚀 View run at https://wandb.ai/team-mykcs/MedSAM-lite/runs/1m08ghs8
  0%|          | 0/906 [00:00<?, ?it/s]Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Finetuning with pretrained weights lite_medsam.pth
MedSAM Lite size: 9791300
Epoch 1 at 2024-04-24 02:47:22, loss: 0.0689: 100%|██████████| 906/906 [13:27<00:00,  1.12it/s]
Traceback (most recent call last):
  File "/Users/myk/PyPjcts/MedSAM_lite-0424/train_mps-1.py", line 535, in <module>
    main()
  File "/Users/myk/PyPjcts/MedSAM_lite-0424/train_mps-1.py", line 515, in main
    "best_loss": best_loss,
UnboundLocalError: local variable 'best_loss' referenced before assignment
wandb: 
wandb: Run history:
wandb:   ce_loss_weight ▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
wandb: epoch_loss[step] ▄▂▁▃▁▄▂█▁▄▁▂▂▆▃▄▅▂▃▃▂▃▃▄▃▁▁▂▂▃▂▂▇▆▄▄▄▄▅▂
wandb:           iou_gt ▁
wandb:  iou_loss_weight ▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
wandb:             l_ce ▂▂▂▁▁▁▂▂▁▃▁▁▂▃▃▁▃▁▃▂▂▁▁▄▁▁▁▂▂▁▁▂█▁▁▂▁▂▃▅
wandb:            l_iou ▂▂▂▁▁▂▁▅▁▂▁▁▁▃▂▃▂▁▁▃▁▂▁▁▁▁▁▂▂▁▂▁█▂▂▁▃▂▃▁
wandb:            l_seg ▄▂▁▃▁▄▂█▁▄▁▂▂▆▃▄▅▂▃▃▂▃▄▄▄▁▁▂▂▃▂▂▅▆▄▄▄▄▅▂
wandb:             loss ▄▂▁▃▁▄▂█▁▄▁▂▂▆▃▄▅▂▃▃▂▃▃▄▃▁▁▂▂▃▂▂▇▆▄▄▄▄▅▂
wandb:      loss.item() ▄▂▁▃▁▄▂█▁▄▁▂▂▆▃▄▅▂▃▃▂▃▃▄▃▁▁▂▂▃▂▂▇▆▄▄▄▄▅▂
wandb:        mask_loss ▄▂▁▃▁▄▂█▁▄▁▂▂▆▃▄▅▂▃▃▂▃▄▄▄▁▁▂▂▃▂▂▆▆▄▄▄▄▅▂
wandb:  seg_loss_weight ▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
wandb: 
wandb: Run summary:
wandb:   ce_loss_weight 1.0
wandb: epoch_loss[step] 0.0689
wandb:           iou_gt 0.84615
wandb:  iou_loss_weight 1.0
wandb:             l_ce 0.00062
wandb:            l_iou 0.00346
wandb:            l_seg 0.06482
wandb:             loss 0.0689
wandb:      loss.item() 0.0689
wandb:        mask_loss 0.06544
wandb:  seg_loss_weight 1.0
wandb: 
wandb: 🚀 View run run-0424-2 at: https://wandb.ai/team-mykcs/MedSAM-lite/runs/1m08ghs8
wandb: ⭐️ View project at: https://wandb.ai/team-mykcs/MedSAM-lite
wandb: Synced 6 W&B file(s), 0 media file(s), 2 artifact file(s) and 0 other file(s)
wandb: Find logs at: ./wandb/run-20240424_023431-1m08ghs8/logs

进程已结束，退出代码为 1
mykcs commented 2 months ago

不是恶性bug，先搁置

mykcs commented 2 months ago

mps-7.py 似乎解决了 推测原因： 把code放入 def main() 里面。 没有细究。