search

fagp / sinkhorn-rebasin

MIT License
15 stars 1 forks source link

Table 3 - how to replicate experiments #10

Open gdacciaro opened 12 months ago

gdacciaro commented 12 months ago

Hi, I really appreciated your work, the paper is clear and well-written. I would need to replicate the experiments (especially on Split_Cifar100), could you then share the hyperparameters or a notebook?

Thank you very much

fagp commented 11 months ago

Hi @gdacciaro, Thank you for your interest in our work and your positive feedback! I have another repo focused on the incremental learning method using the Avalanche library here: https://github.com/fagp/rebasin-IL.

I haven't closed your issue yet because this other repo only includes the RotatedMnist benchmark. However, I'm in the process of cleaning up and refactoring the code for the Split_Cifar100. Once completed, I'll add it to the repository, include a link to this new repo in the readme of sinkhorn-rabasin, and notify you since this seems to be your main interest.

Do not hesitate to reach out if you run into any additional problems.