search

VITA-Group / CADTransformer

[CVPR 2022]"CADTransformer: Panoptic Symbol Spotting Transformer for CAD Drawings", Zhiwen Fan, Tianlong Chen, Peihao Wang, Zhangyang Wang
MIT License
68 stars 20 forks source link

What version of timm do your code use? #6

Closed ziyang1106 closed 1 year ago

ziyang1106 commented 1 year ago

Thanks for your awesome work.

As I run the code CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch --nproc_per_node=1 train_cad_ddp.py --data_root /PATH/TO/PROJECT_DIR/data/FloorPlanCAD --pretrained_model /PATH/TO/PROJECT_DIR/pretrained_models/hrnetv2_w48_imagenet_pretrained.pth some errors occured.

Traceback (most recent call last):
  File "train_cad_ddp.py", line 8, in <module>
    from models.model import CADTransformer
  File "/studio/CADTransformer-main/models/model.py", line 19, in <module>
    from vit import get_vit
  File "/studio/CADTransformer-main/models/vit.py", line 10, in <module>
    from timm.models.vision_transformer import VisionTransformer, _create_vision_transformer, _init_vit_weights, checkpoint_filter_fn, default_cfgs
ImportError: cannot import name '_init_vit_weights' from 'timm.models.vision_transformer' (/opt/conda/envs/cad/lib/python3.7/site-packages/timm/models/vision_transformer.py)

I can't find _init_vit_weights in the latest timm either. The latest version is 0.6.11

Look forward to your reply.

zhiwenfan commented 1 year ago

timm.version '0.4.12'