Comments (4)
Hi Mani,
We also ran into the problem NaNs arising in some training runs.
This problem arises when some learned scaling in the network becomes too large or too close to 0, which then lead an overflow/underflow in the computation.
The easiest thing you can try to avoid this is to decrease the learning rate as we found that this stabilises the training.
An other thing you can experiment with is to bound the learned scalings, we did however not find a good way to do so without decreasing performance.
Let me know if this helps!
Best,
Valentin
from deflow.
Thanks you,
I will try reducing the learning rate first and let you know
Best,
Mani
from deflow.
Reducing the learning rate by a factor of 10 to 0.00001 did the job and I get no more Nans during training.
Thanks a lot
Mani
from deflow.
Great to hear!
from deflow.
Related Issues (11)
- When will the codes be released? HOT 6
- About the cov_shift in the model. HOT 2
- can you share RRDB_PSNR_x16.pth pre-trained model? HOT 1
- Some issues about the NLL loss function
- ModuleNotFoundError: No module named 'utils.ImageSplitter'
- How to train deflow model with pre-trained deflow model you provide by adding own data set? HOT 1
- question about the process of sample
- pretrained models of RRDBnet can't be got HOT 1
- Training on custom dataset
- Applying to low-quality image generation
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 deflow.