search

wusize / raw_ovd

6 stars 0 forks source link

Training COCO and request some files #3

Open qsunyuan opened 1 year ago

qsunyuan commented 1 year ago

command

python train_net.py --num-gpus 8 --config-file configs/test/infer_coco.yaml MODEL.WEIGHTS \ my/path/res50_fpn_soco_star_400.pklpth

bug

model = DistributedDataParallel( File "/opt/conda/envs/pytorch/lib/python3.8/site-packages/torch/nn/parallel/distributed.py", line 648, in init _sync_module_states( File "/opt/conda/envs/pytorch/lib/python3.8/site-packages/torch/distributed/utils.py", line 113, in _sync_module_states _sync_params_and_buffers( File "/opt/conda/envs/pytorch/lib/python3.8/site-packages/torch/distributed/utils.py", line 131, in _sync_params_and_buffers dist._broadcast_coalesced( RuntimeError: The size of tensor a (513) must match the size of tensor b (0) at non-singleton dimension 1

some requests

lvis_v1_train_norare.json and other pkl files.

wusize commented 1 year ago

I will upload them later tonight.

wusize commented 1 year ago

For the bug, use an older version of pytorch like 1.8. Or modify the initialization of quese like the mmdet-based repo.