Reg. Training time - Githubissues

JDAI-CV / image-captioning

Implementation of 'X-Linear Attention Networks for Image Captioning' [CVPR 2020]

268 stars 52 forks source link

Reg. Training time #4

Closed deepak242424 closed 4 years ago

deepak242424 commented 4 years ago

Hi,

Thanks for sharing your code here.

Can you please tell on what type of GPUs you training your model, how much time it took to complete one epoch and the number of epochs till you run your model?

Regards Deepak Mittal

Jordan-5i commented 4 years ago

I run the code on Tesla, 4GPUs, it takes me about 30 min for one epoch with XE.bash experiments/xlan/train.sh

Jordan-5i commented 4 years ago

@deepak242424

Panda-Peter commented 4 years ago

Take xtransformer as an example, 20-30 min for one epoch on 4GPUs (P40) with cross-entropy optimization. The best results are obtained around 30~50 epoches. For RL optimization, it takes ~2.5 hours for one epoch on 1GPU. The best results are obtained around 50 epoches.