search

kjunelee / MetaOptNet

Meta-Learning with Differentiable Convex Optimization (CVPR 2019 Oral)
Apache License 2.0
517 stars 95 forks source link

question about "--episodes-per-batch" #45

Open yxgnahz opened 4 years ago

yxgnahz commented 4 years ago

Hi,

Thanks for your excellent work. Due to the limitation of my GPUs, I couldn't set --episodes-per-batch" to 8 as you did in your paper, instead I set it to 2 and used only one GPU to run your code. However, the result I achieved for miniImagenet 5 way 1 shot accuracy is 59%, which is much lower than your reported result. Could you please tell why "--episodes-per-batch" can influence the result quite significantly?

Thanks

kjunelee commented 4 years ago

You can understand --episodes-per-batch as something equivalent to batch size. Gradient descent is sensitive to the choice of batch size. Instead of reducing --episodes-per-batch, you might want try reducing --train-shot.

By-he commented 4 years ago

Hi,

Thanks for your excellent work. Due to the limitation of my GPUs, I couldn't set --episodes-per-batch" to 8 as you did in your paper, instead I set it to 2 and used only one GPU to run your code. However, the result I achieved for miniImagenet 5 way 1 shot accuracy is 59%, which is much lower than your reported result. Could you please tell why "--episodes-per-batch" can influence the result quite significantly?

Thanks

Hello, I am also reproducing this experiment at present. May I ask that I encountered "TypeError: btrisolve() takes 3 positional arguments but 4 were given" during the operation under the situation of Metaoptnet-RR training on the CIfar-FS。May I ask where this parameter is given more? I am looking forward to your reply. Thank you!