Open jyothsna-phd22 opened 1 month ago
Same here. I am training cfgs/nuscenes_models/cbgs_pp_multihead.yaml
on the nuScenes mini dataset. @jyothsna-phd22 : were you able to resolve the issue?
Tracked down the offending call. It's tv.from_numpy(points)
on this line: https://github.com/open-mmlab/OpenPCDet/blob/master/pcdet/datasets/processor/data_processor.py#L55.
Where should we change in this line??
On Wed, 7 Aug, 2024, 9:32 am Oscar Beijbom, @.***> wrote:
Tracked down the offending call. It's tv.from_numpy(points) on this line: https://github.com/open-mmlab/OpenPCDet/blob/master/pcdet/datasets/processor/data_processor.py#L55 .
— Reply to this email directly, view it on GitHub https://github.com/open-mmlab/OpenPCDet/issues/1642#issuecomment-2272570249, or unsubscribe https://github.com/notifications/unsubscribe-auth/A6SEC5F3VJVK53IPGBH5JALZQGL5BAVCNFSM6AAAAABLZ5O3FCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZSGU3TAMRUHE . You are receiving this because you were mentioned.Message ID: @.***>
I don't yet know how to fix it. I was just sharing that it seems to be the offending line.
On digging further, it seems this is due to a mismatch between the binaries of the spconv library and the rest of the environment. I got it to run using the dockerfile in this PR: https://github.com/open-mmlab/OpenPCDet/pull/1648
This issue is stale because it has been open for 30 days with no activity.
Same issue here,this is my env info:
python 3.10.14
CUDA 12.1
numpy 2.0.1
spconv-cu120 2.3.6
Here is my solution for reference: downgrade numpy version to 1.26.4
conda install numpy=1.26.4
Hi,
While I was training any model on kitti dataset, I am getting floating point exception at epoch 0 and terminated training. How to resolve this issue??
Kindly help me....