You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In your paper 5.2 and table 2, clients with different hardware capacity can use different model architectures.
I wonder how you implement it using pFedHN.
Nice job, thanks!
The text was updated successfully, but these errors were encountered:
Hi, the hypernetworks (HN) outputs a vector of all target model parameters. Then, according to the client's model capacity, we use only the relevant subset of parameters. Concretely, say we have three target models of sizes x1, x2, x3. The HN outputs a vector V of size x1+x2+x3. Now, if the current client has a model of size x2, we use only the x2 parameters V[x2:x3].
We plan on releasing the corresponding code that implements the models in the future.
---Original---
From: "Aviv ***@***.***>
Date: Thu, Jul 1, 2021 21:25 PM
To: ***@***.***>;
Cc: ***@***.******@***.***>;
Subject: Re: [AvivSham/pFedHN] where is the code for part of Computational Budget ? (#2)
Hi, the hypernetworks (HN) outputs a vector of all target model parameters. Then, according to the client's model capacity, we use only the relevant subset of parameters. Concretely, say we have three target models of sizes x1, x2, x3. The HN outputs a vector V of size x1+x2+x3. Now, if the current client has a model of size x2, we use only the x2 parameters V[x2:x3].
We plan on releasing the corresponding code that implements the models in the future.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
In your paper 5.2 and table 2, clients with different hardware capacity can use different model architectures.
I wonder how you implement it using pFedHN.
Nice job, thanks!
The text was updated successfully, but these errors were encountered: