zhnxjtu / CaKDP

(CVPR 2024) CaKDP: Category-aware Knowledge Distillation and Pruning Framework for Lightweight 3D Object Detection
Apache License 2.0
14 stars 1 forks source link

Pruning time consumption #4

Open Li-Whasaka opened 1 month ago

Li-Whasaka commented 1 month ago

Step 2: Prune the models (SECOND as example, on 8 GPUs): How long does it take for you to fall on the dataset once during the pruning iteration process?

zhnxjtu commented 1 month ago

Hi,

Calculating the pruning measure will take some time (not very short), as this step requires iterating over all samples in the dataset. Here, I provide a couple of possible ways to reduce the time consumption:

Additionally, the measure result for SECOND is available at "tools/measure_doc/SECONDNet_kitti.txt".

[1] Lin M, Ji R, Wang Y, et al. Hrank: Filter pruning using high-rank feature map[C], CVPR 2020.

Best regards.