search

lhoyer / DAFormer

[CVPR22] Official Implementation of DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation
Other
470 stars 93 forks source link

python run_experiments.py --exp 101 does not launch the experiments #15

Closed dbash closed 2 years ago

dbash commented 2 years ago

Hi!

I need to run the source-only experiment with your model. However, running python run_experiments.py --exp 101 does not launch the training procedure. Here is what it prints: Run job 220315_1238_zerowaste2zerowaste_source-only_segformer_mitb5_poly10warm_s0_9e0a0 /projectnb2/ivc-ml/dbash/code/DAFormer/mmseg/models/backbones/mix_transformer.py:214: UserWarning: DeprecationWarning: pretrained is a deprecated, please use "init_cfg" instead warnings.warn('DeprecationWarning: pretrained is a deprecated, ' /projectnb/ivc-ml/dbash/.conda/envs/daformer/lib/python3.8/site-packages/mmcv/cnn/utils/weight_init.py:118: UserWarning: init_cfg without layer key, if you do not define override key either, this init_cfg will do nothing warnings.warn(

After that, it just exits. I was wondering if I did something wrong here and if you could help me with this issue. The only changes I've made is I commented out L421-428 in experiments.py.

Thank you for your time!

lhoyer commented 2 years ago

Hi Dina,

Experiment 101 is only supposed to test if all configurations launch properly. Therefore, it is set to only train for two iterations and to only report errors. The behavior you are describing is intended.

To run a proper training, please use experiment 7 for example. To train source-only, you can just set uda = 'source-only' in line 369.

Best, Lukas

dbash commented 2 years ago

Thank you for a quick response!