New year, new docker images! This time, we have refreshed our docker images that use libaries from the OpenMMLab group:
MMClassification 0.25.0 (CPU and CUDA 11.1)
MMSegmentation 0.30.0 (CPU and CUDA 11.1)
MMDetection 2.27.0 (CPU and CUDA 11.1)
All frameworks now offer a script (mmcls/mmseg/mmdet_onnx) to export a PyTorch model to ONNX.
MMDetection now also allows you to select the CUDA device to train on rather than just always using the first available GPU.