Comments (12)
Still training……
QQ
from hyperkohaku.
It should be usable
But "how to get reasonable result" still need more research
from hyperkohaku.
Still training……
@KohakuBlueleaf are you still training the loras or are you already training HyperDreamBooth?
from hyperkohaku.
Still training……
@KohakuBlueleaf are you still training the loras or are you already training HyperDreamBooth?
Hyperdreambooth
You can actually skip stage I
Directly train the hypernetwork is doable
Already tried on LyCORIS repo
The problem is I need someone to help me to implement something like "mask face" trick for train on face
Or something like that
I have not much experience about DreamBooth so cannot get a good training method for pre-optimized loras
from hyperkohaku.
If you skip stage 1, what do you use as ground truth weights during the hypernetwork training (Figure 2, phase 1 in the paper)? Do you randomly initialize them? I am trying to understand it from your code but I am a bit lost
from hyperkohaku.
If you skip stage 1, what do you use as ground truth weights during the hypernetwork training (Figure 2, phase 1 in the paper)? Do you randomly initialize them? I am trying to understand it from your code but I am a bit lost
Actually you only need the diffusion loss
Weight loss is jot necessary
But it will give you better ( if your pre-optimized weights is good) result
from hyperkohaku.
@KohakuBlueleaf If the stage 1 is skipped, should we set the add_constant flag to True in the HyperDream?
from hyperkohaku.
@KohakuBlueleaf If the stage 1 is skipped, should we set the add_constant flag to True in the HyperDream?
I think yes
But I am not sure if it will work
Will check my code
from hyperkohaku.
@KohakuBlueleaf Hi is this project stop now?
from hyperkohaku.
@KohakuBlueleaf Hi is this project stop now?
Not, but I just don't have time to let it keep moving...
I have literally over 10 proj on hand....
from hyperkohaku.
Thank you for your hard work. Is the hyperdreambooth project a low priority among the 10 ongoing projects? @KohakuBlueleaf
from hyperkohaku.
Thank you for your hard work. Is the hyperdreambooth project a low priority among the 10 ongoing projects? @KohakuBlueleaf
Yes it is relatively low
from hyperkohaku.
Related Issues (19)
- i would like to contribute in this project HOT 1
- Pretrained hyper network weights HOT 2
- Section 4: further finetuning HOT 5
- Read Me
- celeba-hq-512-4.8k dataset HOT 1
- environment problem HOT 2
- Multi GPU parallelization HOT 4
- safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge
- questions about running requirements HOT 3
- Multi-gpu training cannot be performed
- How to mask the face during training? HOT 4
- Loss does not converge during HyperNet training HOT 3
- Pre Optimize problem HOT 6
- Pretrained HyperNetwork model HOT 5
- Question about the LiLora aux matrix HOT 31
- Is there anyone who succeeded? HOT 11
- I cant' find where weight_preds are loaded into T2I model HOT 3
- diffusers version HOT 2
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 hyperkohaku.