fengyue20
Results
3
comments of
fengyue20
I have this issue too. Did you change the {CONFIG_FILE} to 'configs/MEGA/vid_R_101_C4_MEGA_1x.yaml'
应该不支持，因为用了retinanet就无法用ROI-Align, 无法build_roi_heads，因此detector_losses = {}。没有detector_losses还怎么训练
当前环境下pip升级了transformer，但是paddle里的Transformer不变，怎么办。微调成功了，但是微调后和原模型融合不了，没法用微调的模型推理啊