Jonas
@jonasloos.bsky.social
ai/ml, diffusion, xai, llms, compilers, robotics, space
Alright, thanks for the answer 👍
December 9, 2024 at 12:24 PM
Alright, thanks for the answer 👍
Very nice work!
Why do you do the ablation studies only on a subset of the testset of Spair-71k, is it due to compute? In my experiments, running sc on the full spair test split takes less than 10min on a single 3090 🤔
Why do you do the ablation studies only on a subset of the testset of Spair-71k, is it due to compute? In my experiments, running sc on the full spair test split takes less than 10min on a single 3090 🤔
December 7, 2024 at 1:24 PM
Very nice work!
Why do you do the ablation studies only on a subset of the testset of Spair-71k, is it due to compute? In my experiments, running sc on the full spair test split takes less than 10min on a single 3090 🤔
Why do you do the ablation studies only on a subset of the testset of Spair-71k, is it due to compute? In my experiments, running sc on the full spair test split takes less than 10min on a single 3090 🤔
Yeah. The idea of entangling channels seems like someone might have tried it already (in another context). But I didn't immediately find something
November 24, 2024 at 5:18 PM
Yeah. The idea of entangling channels seems like someone might have tried it already (in another context). But I didn't immediately find something
Do you have an example where entangling the channels works better than not entangling them?
November 24, 2024 at 4:57 PM
Do you have an example where entangling the channels works better than not entangling them?
Nice idea to preserve additional information! But why the complex plane and not some arbitrary number of extra channels per pixel?
November 24, 2024 at 3:08 AM
Nice idea to preserve additional information! But why the complex plane and not some arbitrary number of extra channels per pixel?