shi-labs / forget-me-not Goto Github PK
View Code? Open in Web Editor NEWForget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models, 2023
License: MIT License
Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models, 2023
License: MIT License
Hi. Thanks for the great paper. I wanted to replicate the results for the art style of Van Gogh. It would be really helpful if you could answer the questions
ti.yaml
, can you give more information on how to change the placeholder tokens? Did you tune Ti
in your experiments for art styles in the paper?RuntimeError: torch.cat(): expected a non-empty list of Tensors
at line 287 in train_attn.py. Could you please let me know how to solve this?Hello and thank you for the great work!
There was already a question regarding the code release date. Do you have any plans to share the ConceptBench dataset. If so, when can we expect that? Thank you!
Hi,
I recently went through the paper and really liked your work!!!
I went through Section 4.4 and found the evaluation metric concept bench very interesting but I didn't find any code in the github repo for getting memorization scores of a concept forgotten, could you please add in the code as well or give any direction to how memorization scores were calculated effectively especially this step : "Subsequently, we invert the same anchor images of Elon Musk using original model and forgetting model respectively." ?
Thanks,
Kartik
Thanks for your excellent works!
I've try your example of Elon Musk and the result is amazing, I try to use the same setting for NSFW erasing but it seems not working.
Can you shares your setting for forgetting prompt "a photo of naked"? Thank you.
Hi, thans for the wonderful work!
I'm currently try to use FMN to removel multipile concepts with only one model to train. However in your instructions and example configs, only single concept is mentioned and initializer_tokens
in ti.yaml
seem to highly dependent on the concepts.
I'm wondering what modification should I made to configurate them correctly for multi-concept forgetting.
Hi,
I tried to find the proper place to run inference after the deletion of some concepts. I wanted to infer something that is not identical to the original conecpt that was deleted. How can I do that?
Thanks!
Hi! Thanks for your great work! When will you release the code? Thanks!
Hi authors, I was recently going through your paper, Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models, and wanted to replicate the experimental results as state in the paper. To this end, I followed the necessary steps for setup, and thereafter, execute the first on compilation of the code as shown,
python run.py configs/ti.yaml
Thereafter, on running the second stage of compilation
python run.py configs/attn.yaml
we encountered the following Runtime Error
RuntimeError: torch.cat(): expected a non-empty list of Tensor
Could you please help us in fixing this problem
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.