Barry, Y. Ding

Hi @EternalSaga, Thanks for the script. I tried to put your commands into a bash script like: ``` SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd...

I found out that I forgot to prepare spconv/CMakeLists.txt. It works like a charm with this [repo's example](https://github.com/traveller59/spconv/blob/master/example/libspconv/spconv/CMakeLists.txt): ``` cmake_minimum_required(VERSION 3.20) project(spconv LANGUAGES CXX CUDA) include_directories(/builds/libspconv/include) add_subdirectory(src) # tell parent...

In addition, reverting this swapping, in other words, init-ing the heatmap as shape _feature_map_size[0], feature_map_size[1]_ can accomplish to train a model.

> Are you training a camera-only model? No, I’m dealing with Lidar-only and LC-fusion models. But this bug will remain there with cam-only BEVFusion, as the vtransform outputs a Bev...

> In fact, this operation is present in the original code of transfusion head in BEVFusion. However, the mit-bevfusion and the BEVFusion from NeurIPS 2022 differ in the final step...

> Have you tried a non-square bev LCFusion? Does the 2d backbone of the bev model accept input properly? Yes, that's exactly the case I'm encountering. I have sparse_shape=[960, 1088,...

@cxnaive Actually there is another confusing snippet, which however might be a hint to understand these ambiguous spatial shapes: https://github.com/open-mmlab/mmdetection3d/blob/fe25f7a51d36e3702f961e198894580d83c4387b/projects/BEVFusion/bevfusion/transfusion_head.py#L727-L731 Say sparse_shape (xyz) is [960, 1088 41]. The predicted center...

> https://github.com/open-mmlab/mmdetection3d/blob/fe25f7a51d36e3702f961e198894580d83c4387b/mmdet3d/models/utils/gaussian.py#L46C5-L53C69 From this, it can be seen that the default heatmap shape in mmdet3d is [Y, X] This solves most of the confusion. That's why initializing the heatmap as...

Hi @hopef, Thank you for the clarification. It's a pity missing the chance to inspect the techniques used to achieve such significant acceleration. Nevertheless, may I learn about what could...

@hopef Thank you so much for the hints (while sorry for the late response). I managed to reuse the rulebook for every 4 consecutive sub-manifold conv layers and indeed achieved...