RuntimeError: cuda runtime error (77) : an illegal memory access was encountered at /data/parsing/models/ops/csrc/cuda/PoolPointsInterp_cuda.cu:223

[arielsuee11]

firstly, I encountered this problem

Traceback (most recent call last): File "tools/train_net.py", line 147, in main() File "tools/train_net.py", line 142, in main train(model, train_loader, optimizer, scheduler, checkpointer, training_logger) File "tools/train_net.py", line 73, in train outputs = model(images, targets) File "/data/parsing_env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, **kwargs) File "/data/parsing_env/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py", line 150, in forward return self.module(*inputs[0], **kwargs[0]) File "/data/parsing_env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, **kwargs) File "/data/parsing/rcnn/modeling/model_builder.py", line 138, in forward x, result, loss_uv = self.UV_RCNN(conv_features, result, targets) File "/data/parsing_env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, **kwargs) File "/data/parsing/rcnn/modeling/uv_rcnn/uv_rcnn.py", line 31, in forward return self._forward_train(conv_features, proposals, targets) File "/data/parsing/rcnn/modeling/uv_rcnn/uv_rcnn.py", line 43, in _forward_train loss_Upoints, loss_Vpoints, loss_seg_AnnIndex, loss_IndexUVPoints = self.loss_evaluator(uv_logits) File "/data/parsing/rcnn/modeling/uv_rcnn/loss.py", line 155, in call interp_U = PPI_op(UV_pred_U, Coordinate_Shapes) File "/data/parsing_env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in call result = self.forward(*input, **kwargs) File "/data/parsing_env/lib/python3.6/site-packages/apex-0.1-py3.6-linux-x86_64.egg/apex/amp/amp.py", line 22, in wrapper return orig_fn(*args, **kwargs) File "/data/parsing/models/ops/pool_points_interp.py", line 59, in forward return pool_points_interp(input, rois, self.spatial_scale) File "/data/parsing/models/ops/pool_points_interp.py", line 27, in forward input, roi, spatial_scale) RuntimeError: cuda runtime error (77) : an illegal memory access was encountered at /data/parsing/models/ops/csrc/cuda/PoolPointsInterp_cuda.cu:223

it seems that the three variables (input,roi,spatial_scale) are not in cuda, so I change the spatial_scale into torch.float.tensor ，but it seems useless. So hope you someone can help me solve this problem,thanks a lot!