FlyingQianMM

[Feature] get maximum device memory allocated in train log

2

In training log, maximum device memory allocated is shown.

reset save_intervals default value in toos/train.py: - if users use `iter` during the traing phase, save_intervals default value is set to `1000` - if users use `epochs` during the traing...

To compute kitti eval result, the 3d boxes on the lidar coordinate need to be transformed to the one on the kitti camera coodinate. On the kitti camera coordinate, the...

Add 3d sparse conv for centerpoint: - middle_encoder is composed with a sparse 3D resnet - accelerate `get_voxel_idx_kernel ` in the hard voxelize custom operation

train dataset would not been initialized during evaluation

2

train dataset would not been initialized during evaluationtrain dataset would not been initializer during evaluation

[WIP][Bug fix] use pad_shape instead of img_shape for bevformer and recover the origin code where the first frame pre_bev is None

2

- Use `pad_shape` instead of `img_shape` for bevformer, where `img_shape` reserves the origin image shape and `pad_shape` is the image shape after padded. - Recover the origin codes where the...

fix the bug where convert bn to sync_bn in the model for evaluation during the training phase will change the training model graph

1

The origin implementation converts bn to sync_bn in the model for evaluation during the multi-gpu training phase, which is not necessarily and changes the training model graph, causes the gradient...

[New feature] add dynamic voxelizer implementation

2

- Add dynamic voxelizer implementation - Add a training config with the dynamic voxelizer for centerpoint

paddlex gui 2.0.0rc1 windows运行时间统计问题

2

 才开始训练一会儿，显示运行了8个多小时，还剩下5个小时，而且这个时间随机乱跳，用的是ppyolotiny。