Unstable results

[CuthbertCai]

Hi, thanks for your contribution!

I've re-run the model twice with GCC+RedCaps. For the first time, I got 51.9 on ImageNet and 11.6 on COCO, while for the second time, I got 54.3 on ImageNet and 25.8 on COCO. It seems that the results are not stable.

Is this a normal phenomenon for this model?

[xvjiarui]

Hi @CuthbertCai

Sorry for the late reply.

Are you running training or inference? If you are running training, fixing seed will make the result more stable.

[bryanyzhu]

@CuthbertCai Hi, do you have results on VOC? Is it stable? We used the GCC+RedCaps config, and got very different results for different runs. Does fixing seed help the stableness? Thank you.