Comments (13)
@Crawlling Did you specify the correct time steps (t) here corresponding to the fine-tuned model you use in the test config:
ddpm-cd/config/dsifn_test.json
Line 63 in 4876ec6
from ddpm-cd.
Are you using the correct pre-trained CD model? There are different configurations (t).
from ddpm-cd.
Are you using the correct pre-trained CD model? There are different configurations (t).
thx for reply!i've check the pre-trained CD model and config,it is correct,i don't know why this alway happen,i've already try it on 2 PC.
but i try to use pre-trained Diffusion model to train CD head model,in epoch 0,it still got pretty nice F1 and iou.thank you.
from ddpm-cd.
I see. I will look into this problem. The thing is, it is working fine with my machine. I am happy that at least you managed to fine-tune it and got reasonable iou. I will look into this problem and update you if I find the reason. In case if you find the reason, pls post it here so that it will benefit others. Thank you.
from ddpm-cd.
@Crawlling np.
from ddpm-cd.
help please
from ddpm-cd.
All the pretrained CD head models i can not use..but i can use the diffusion model.
from ddpm-cd.
I see. I will look into this problem. The thing is, it is working fine with my machine. I am happy that at least you managed to fine-tune it and got reasonable iou. I will look into this problem and update you if I find the reason. In case if you find the reason, pls post it here so that it will benefit others. Thank you.
thank you for your patience,i will.
from ddpm-cd.
specify
OMG,i forgot this,my bad, sorry for wasting your time.thank you again!
from ddpm-cd.
How much memory does this method take? Is 24GB not enough by default?
from ddpm-cd.
Hi @sitongzhen
Do you mean pre-training or fine-tuning?
I used one 48GB GPU for my experiments, and I believe a 24GB GPU should be fine. You just need to reduce the batch size in the config file. However, you can significantly save time by initializing your model with the pre-trained models provided here, as that will make the training faster even on new datasets.
Feel free to reach out to me know if you have more questions.
Also, the code works on multiple GPUs as well.
from ddpm-cd.
Yes, I understand. Thanks for your answer. When the batch is 8, I find the 24 G memory is insufficient in the training process.
from ddpm-cd.
@sitongzhen I don't expect results to vary significantly, and it is fine given the memory constraint.
I see, make it down to 4 and see the results.
Good luck!
from ddpm-cd.
Related Issues (20)
- About can't load the trained model. HOT 2
- has a wrong answer
- Training diffusion model with remote sensing data HOT 2
- Data for pretraining the diffusion model HOT 1
- question for 'GaussianDiffusion'
- About the expansion of the pre-training dataset HOT 2
- Question for typo in paper HOT 1
- Prediction Images Are All Black HOT 3
- Practicality of the model on 64x64 images HOT 7
- TypeError: unsupported operand type(s) for %: 'int' and 'NoneType' HOT 1
- When loading your pretraining file, errors may occur in the following model file sections due to structural mismatch between your pretraining model directly under the link and the ddpm model under ddpm_cd. HOT 4
- Broken wandb link HOT 1
- Train/Val Reports on wandb can not be opened HOT 4
- how to reproduce Figure3 in paper HOT 2
- dppm
- How to download the unlabel datasets?
- an issue occured when using the pretrained weight
- From the perspective of diffusion model principle, why are the representations generated by DDPM from remote sensing images more robust and distinguishable than those obtained by UNet networks? HOT 2
- RuntimeError: Error(s) in loading state_dict for GaussianDiffusion: HOT 1
- about _opt.pth and _gen.pth
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 ddpm-cd.