When model's pretrained is none, get AttributeError: 'ViTAE_Window_NoShift_basic' object has no attribute 'use_abs_pos_embed' about rsp HOT 4 CLOSED

@AnRanbel please show the whole error report

from rsp.

@AnRanbel please show the whole error report

谢谢你的回复，我有将你们的代码移植到最新版的mmsegmentation里面，所以有些路径可能不同，但我训练和测试过了，除了不设置pretrained的时候会报错，其他都是正常运行的。

Exception has occurred: AttributeError (note: full exception trace is shown but execution is paused at: _run_module_as_main)
'ViTAE_Window_NoShift_basic' object has no attribute 'use_abs_pos_embed'
File "/root/miniconda3/envs/mmyolo/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
File "/root/autodl-tmp/mmsegmentation/mmseg/models/backbones/ViTAE_Window_NoShift/base_model.py", line 197, in init_weights
if self.use_abs_pos_embed:
File "/root/miniconda3/envs/mmyolo/lib/python3.8/site-packages/mmengine/model/base_module.py", line 130, in init_weights
m.init_weights()
File "/root/miniconda3/envs/mmyolo/lib/python3.8/site-packages/mmengine/runner/runner.py", line 899, in _init_model_weights
model.init_weights()
File "/root/miniconda3/envs/mmyolo/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1707, in train
self._init_model_weights()
File "/root/autodl-tmp/mmsegmentation/tools/train.py", line 134, in main
runner.train()
File "/root/autodl-tmp/mmsegmentation/tools/train.py", line 138, in
main()
File "/root/miniconda3/envs/mmyolo/lib/python3.8/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/root/miniconda3/envs/mmyolo/lib/python3.8/runpy.py", line 192, in _run_module_as_main (Current frame)
return _run_code(code, main_globals, None,
AttributeError: 'ViTAE_Window_NoShift_basic' object has no attribute 'use_abs_pos_embed'

from rsp.

@AnRanbel ViTAEv2中没有绝对位置编码，这段加载权重的代码可能是从什么地方直接复制过来的，当时没考虑不加载权重的情况，所以pretrained=None的时候，你把它注释掉

# if self.use_abs_pos_embed:
#    trunc_normal_(self.absolute_pos_embed, std=0.02)

from rsp.

@AnRanbel ViTAEv2中没有绝对位置编码，这段加载权重的代码可能是从什么地方直接复制过来的，当时没考虑不加载权重的情况，所以pretrained=None的时候，你把它注释掉

# if self.use_abs_pos_embed:
#    trunc_normal_(self.absolute_pos_embed, std=0.02)

好的，谢谢。

from rsp.