-
Notifications
You must be signed in to change notification settings - Fork 368
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Bug] the number of embedddings in ManagedCollisionCollection must be a multiple of the number of devices #1591
Comments
Hi, thanks for trying out ManagedCollisionCollection! Not sure if its a bug. The thing is, we are trying to (only) use ManagedCollisionCollection with rowwise sharding, which would shard the table evenly to all the gpus, hence the divisible thing. |
Thanks for your quick response, yes, I tried ManagedCollisionCollection on our data, the performance degraded when using ManagedCollisionCollection. The training time is also significant increased. Is there any guideline or document on how to set the hyper-parameters when using this module, e.g., eviction_interval, zch_size, mch_size, and which policy is better DistanceLFU_EvictionPolicy or LFU_EvictionPolicy under which scenario. |
@fangleigit Thanks. We are still actively developing MCH/ZCH, so we don't have a clear answer so far. Let us know if you have it figured out as well! |
when changing the number of embeddings to 4091, and mch_size to 1021 of the code below, it will throw the following exception
The text was updated successfully, but these errors were encountered: