when run demo: RuntimeError: CUDA error: unknown error

[littlejiumi]

Traceback (most recent call last): File "tools/demo.py", line 190, in main() File "tools/demo.py", line 131, in main cfg.model, train_cfg=cfg.train_cfg, test_cfg=cfg.test_cfg) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 60, in build_detector return build(cfg, DETECTORS, dict(train_cfg=train_cfg, test_cfg=test_cfg)) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 32, in build return _build_module(cfg, registry, default_args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 24, in _build_module return obj_type(**args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/detectors/smpl_rcnn.py", line 65, in init self.smpl_head = builder.build_head(smpl_head) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 52, in build_head return build(cfg, HEADS) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 32, in build return _build_module(cfg, registry, default_args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 24, in _build_module return obj_type(**args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/smpl_heads/smpl_head.py", line 74, in init self.loss = build_loss(loss_cfg) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 56, in build_loss return build(cfg, LOSSES) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 32, in build return _build_module(cfg, registry, default_args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/builder.py", line 24, in _build_module return obj_type(**args) File "/root/picasso/xhzhang/multiperson/mmdetection/mmdet/models/losses/smpl_loss.py", line 107, in init anti_aliasing=False) File "/root/anaconda3/envs/multiperson/lib/python3.7/site-packages/neural_renderer_pytorch-1.1.3-py3.7-linux-x86_64.egg/neural_renderer/renderer.py", line 41, in init self.dist_coeffs = torch.cuda.FloatTensor([[0., 0., 0., 0., 0.]]) RuntimeError: CUDA error: unknown error

[JiangWenPL]

I didn't meet this issue in our environment.