DAT-Detection icon indicating copy to clipboard operation
DAT-Detection copied to clipboard

Please tell me how to fix this error

Open Fight4fun-ych opened this issue 2 years ago • 4 comments

Now i have the new questions as follows

2023-11-08 03:17:02,487 - mmdet - INFO - Set random seed to 67311741, deterministic: False Traceback (most recent call last): File "/home/ych/DAT-Detection/DAT-Detection-main/tools/train.py", line 247, in main() File "/home/ych/DAT-Detection/DAT-Detection-main/tools/train.py", line 219, in main model.init_weights() File "/home/ych/.conda/envs/python39/lib/python3.9/site-packages/mmcv/runner/base_module.py", line 117, in init_weights m.init_weights() File "/home/ych/DAT-Detection/DAT-Detection-main/models/backbones/dat.py", line 332, in init_weights load_checkpoint(self, pretrained, strict=False, logger=logger) File "/home/ych/DAT-Detection/DAT-Detection-main/models/backbones/loading.py", line 362, in load_checkpoint table_current = model.state_dict()[k1] KeyError: 'backbone.stages.0.attns.1.rpe_table'

when i try "python tools/train.py configs/dat/rtn_tiny_1x_4n_dp00_lr2.py" and i make sure i have update the pretrain variable in <CONFIG_FILE> as base = [ '../base/models/retinanet_dat_fpn.py', '../base/datasets/coco_detection.py', '../base/schedules/schedule_1x.py', '../base/default_runtime.py' ]

pretrained = '/home/ych/DAT-Detection/DAT-Detection-main/rtn_dat_t_1x.pth'

could somebody help me????

Fight4fun-ych avatar Nov 02 '23 09:11 Fight4fun-ych

I am having the same problem. Any help would be appreciated!!

anzr299 avatar Mar 17 '24 08:03 anzr299

Did you fix the problem?

anzr299 avatar Mar 17 '24 09:03 anzr299

I am having the same problem. Any help would be appreciated!!

yes,backbone should use dat_pp_tiny_in1k_224.pth to initialize,download by classfication mission

Fight4fun-ych avatar Mar 18 '24 01:03 Fight4fun-ych

I am having the same problem. Any help would be appreciated!!

yes,backbone should use dat_pp_tiny_in1k_224.pth to initialize,download by classfication mission

Should I continue using pretrained or should I use load_from? Also what do you mean by the classification mission?

anzr299 avatar Mar 18 '24 06:03 anzr299