Closed litingsjj closed 7 months ago
Please use the latest pytorch, at least the pytorch version should be > 2.0.0 to support sparse matrix multiplication.
Got it!
here is my result, the miou is different from yours
InverseMatrixVT3D_200.pth:
02/08 18:44:31 - mmengine - INFO - Epoch(test) [753/753] noise: nan barrier: 0.1902 bicycle: 0.1439 bus: 0.4121 car: 0.4006 construction_vehicle: 0.1298 motorcycle: 0.2227 pedestrian: 0.1982 traffic_cone: 0.1233 trailer: 0.1877 truck: 0.3154 driveable_surface: 0.4026 other_flat: 0.1491 sidewalk: 0.1972 terrain: 0.2272 manmade: 0.1661 vegetation: 0.2890 miou: 0.2347 acc: 0.2999 acc_cls: 0.2888 data_time: 0.0061 time: 0.3000
InverseMatrixVT3D_256.pth:
02/08 18:55:23 - mmengine - INFO - Epoch(test) [753/753] noise: nan barrier: 0.2065 bicycle: 0.1313 bus: 0.4340 car: 0.3646 construction_vehicle: 0.1319 motorcycle: 0.2254 pedestrian: 0.2064 traffic_cone: 0.1162 trailer: 0.2377 truck: 0.3468 driveable_surface: 0.3475 other_flat: 0.1390 sidewalk: 0.2060 terrain: 0.2075 manmade: 0.1659 vegetation: 0.2632 miou: 0.2331 acc: 0.2747 acc_cls: 0.2959 data_time: 0.0057 time: 0.3710
please try set the enable_fix=False or Ture during test, the released code has been modified a lot but the ckpt files are old. You can also try train from scratch, the best performance (mIoU goes to 26%) achieved after around 4 epoches.
Please set enable_fix=False in each config file and try testing the model again.
Solved it with enable_fix=False
hi, it's a great work! i got an error, when i run
bash tools/dist_test.sh configs/InverseMatrixVT3D_200_200_16.py InverseMatrixVT3D_200.pth 8
NotImplementedError: Could not run 'aten::sum.IntList_out' with arguments from the 'SparseCUDA' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using cu stom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'aten::sum.IntList_out' is only available for these backends: [Dense, FPGA, Conjugate, UNKNOWN_TENSOR_TYPE _ID, QuantizedXPU, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, SparseCPU, SparseCUDA, SparseHIP, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ ID, SparseXPU, UNKNOWN_TENSOR_TYPE_ID, SparseVE, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, NestedTensorCUDA, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOW N_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID].