Comments (4)
works now, thank you 👍
I open-sourced my code for experimenting with AdaHessian where I used your optimizer implementation (of course added copyright + link to your repo).
from ada-hessian.
Hi, you are absolutely right. While correct accumulation could be supported, it seems that it would make the code much more complicated and it would require more memory (for storing both and ), so I'll completely remove this feature in the next version. It doesn't seem to be worth it for the default (more common) simple usage.
On the other hand, better approximation of formula 11 doesn't seem to need accumulated gradients, all samples of should be computed from the same gradient. I'll add an option to use more samples :)
from ada-hessian.
Yes, I also did it that way: simply pass number of samples to set_hessian(num_samples), accumulate z*(H@z) values, and finally divide by that number. But good to know that this will be in the next release :-)
from ada-hessian.
That's amazing, thank you for sharing the experiments!
from ada-hessian.
Related Issues (6)
- Comparison with ref. implementation and small issue with namespace HOT 4
- torch.autograd.grad(grads, params, grad_outputs=z, only_inputs=True, retain_graph=False) local variable 'z' referenced before assignment HOT 2
- AttributeError: 'Parameter' object has no attribute 'hess' HOT 3
- question: about create_graph HOT 1
- bug about generater HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ada-hessian.