How many batch sizes are there? And how many GPUs are used?
I use one A40 GPU, set the batch size to 128, and set the learning rate to 0.0004 (All changes in file "B4_cae_center_got10k_ep100.yaml").
But the performance of the tracker is: "AO: 61.8". It is significantly smaller than 65.2 (reported in your paper). The pretrained model is "cae_base_1600ep.pth".
How many batch sizes are there? And how many GPUs are used?
I use one A40 GPU, set the batch size to 128, and set the learning rate to 0.0004 (All changes in file "B4_cae_center_got10k_ep100.yaml").
But the performance of the tracker is: "AO: 61.8". It is significantly smaller than 65.2 (reported in your paper). The pretrained model is "cae_base_1600ep.pth".