HaoTian
HaoTian
I'm not sure about your question. If you want to make the similiar demo in readme, maybe ref this file https://github.com/fundamentalvision/BEVFormer/blob/master/tools/analysis_tools/visual.py
In v2, Its implemented by align the cam2lidar transformation matrix, maybe called online camera on many other framework. See here https://github.com/fundamentalvision/BEVFormer/blob/32b65ffd2a85c35b5f74b753f23ef34f9a85a576/projects/mmdet3d_plugin/datasets/nuscenes_dataset_v2.py#L137 In v1, Its implemented by transfer the prev bev...
It seems your IPython version is not correct.
For Motion prediction and then for planning. See [UniAD ](https://arxiv.org/abs/2212.10156) to have a general sense of the entire system.
Setting use_can_bus=False. It should remove the can_bus embedding layer.
Check here: https://github.com/OpenGVLab/InternVL/blob/main/BLOG.md#data-preparation
For flash attention in windows, please follow the Instruction in https://github.com/Dao-AILab/flash-attention. It might be tough to run with flash attention in windows.
Firstly whether the provided ckpt is able to get exactly same performance in your env? The training log are also provided, with the same setting, check the whether are exactly...
https://huggingface.co/spaces/Qwen/Qwen2-VL. They say they might release the model within Aug in X.