#track 2: 'Solutions can exploit a replay buffer containing data coming from up to 3500 training instances.' For example, in task1, can we randomly sample 3500 images from the full training set (ego_objects_challenge_train.json)?

[Layjins]

#track 2: 'Solutions can exploit a replay buffer containing data coming from up to 3500 training instances.' For example, in task1, can we randomly sample 3500 images from the full training set (ego_objects_challenge_train.json)?

[lrzpellegrini]

Hi @Layjins, the replay buffer can be filled only using data from the current experience. You can't use data from future or past experiences. This also means that the initial replay buffer must be empty!

You will have to choose which images to keep and which to discard before completing each training experience (this choice can greatly impact the final performance of your solution).

[Layjins]

thanks a lot!

…

------------------ Original ------------------ From: Lorenzo Pellegrini ***@***.***> Date: Tue, Apr 19, 2022 11:07 PM To: ContinualAI/clvision-challenge-2022 ***@***.***> Cc: Layjuns ***@***.***>, Mention ***@***.***> Subject: Re: [ContinualAI/clvision-challenge-2022] #track 2: 'Solutions canexploit a replay buffer containing data coming from up to 3500 traininginstances.' For example, in task1, can we randomly sample 3500 images from thefull training set (ego_objects_challenge_train.json)? (Issue #13) Hi @Layjins, the replay buffer can be filled only using data from the current experience. You can't use data from future or past experiences. This also means that the initial replay buffer must be empty! You will have to choose which images to keep and which to discard before completing each training experience (this choice can greatly impact the final performance of your solution). — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>