Closed zml110120 closed 2 years ago
V100-32GB, batch size of 10, single GPU is enough.
Actually 16G is enough, I try to train under SCST with batch size of 10, it just cost 5000M - 6000M CUDA memory.
Actually 16G is enough, I try to train under SCST with batch size of 10, it just cost 5000M - 6000M CUDA memory.
Oh, I am sorry, it is wrong, I trained the model without the backbone model of SwinT, the full model may need more CUDA memory.
Could you tell me which kind of GPU and how many you has uesd? I use V100-16GB and there is OOM error