Roy @roymiles.bsky.social · 09/12/2024Hey! I will be at NeurIPS @neuripsconf.bsky.social this week (friday 13th) presenting our paper VeLoRA. tldr; simple method for compressing gradients with a coarse reconstruction during the backwards pass. Significant memory reductions while being complimentary to LoRA! github.com/roymiles/VeL... 160
Roy @roymiles.bsky.social · 27/11/2024Haha I wasn't really thinking about it like that. In either case I have added you to this one too. 020
Roy @roymiles.bsky.social · 27/11/2024Hey, I have added you! oops yea I did see that one. Perhaps this one can be a bit more broader with other modalities. 220
Roy @roymiles.bsky.social · 27/11/2024Hey! Yea yours probably overlaps a lot 😅 I have added you and will add some others from there too. 010
Roy @roymiles.bsky.social · 27/11/2024I made a starter pack for people working or interested in multi-modality learning. It would be good to add lots more people so do comment and I'll add! go.bsky.app/97fAH2N 15206
Reposted by RoyBenedikt Kolbeinsson @handle.invalid · 27/11/2024Our BMVC poster is today! Knowledge distillation can work in cross-task settings and even with a randomly initialised teacher! Our inverted projection decomposes into knowledge transfer and spectral regularisation, enabling teacher-free distillation with improvements on ImageNet-1K at no extra costarxiv.org 073
Roy @roymiles.bsky.social · 23/11/2024Hey, this is great and relevant for me. Would love to join if there is still space! 110
Roy @roymiles.bsky.social · 22/11/2024This is really great thanks! Could I be added to the AI starter pack too 😅? 020
Roy @roymiles.bsky.social · 22/11/2024Some related work worth checking out if you find this cool. mbaradad.github.io/shaders21k/ - learning good visual features from procedurally generated images. arxiv.org/abs/2403.14494 - distillation from randomly weighted teachers. 020
Roy @roymiles.bsky.social · 21/11/2024Hi Bluesky! I'm currently working on multi-modality learning and efficient ML (mobile devices). Hoping to regularly post about any interesting topics I find cool (ai, maths, physics). Looking forward to seeing how this platforms grows and meeting you all! 130