The annotated PyTorch training loopidlemachines.co.uk81 points·by smaddrellmander·20 ngày trước·19 comments
Mae vs. MSE: more than just the mean vs. median debateidlemachines.co.uk1 points·by smaddrellmander·24 ngày trước·0 comments
DiffusionGemma: Discrete diffusion in a large language modelidlemachines.co.uk3 points·by smaddrellmander·tháng trước·1 comments
Reading MAI's efficiency gain. How to pick architectures like serious peopleidlemachines.co.uk9 points·by smaddrellmander·tháng trước·0 comments
MAI-Thinking-1: Building a Hill-Climbing Machine [pdf]microsoft.ai3 points·by smaddrellmander·tháng trước·0 comments
Are contrastive losses just cross entropy all along?idlemachines.co.uk2 points·by smaddrellmander·tháng trước·0 comments
Every token, everywhere, all at onceidlemachines.co.uk2 points·by smaddrellmander·2 tháng trước·0 comments
The cut in the Mixture of Experts compute graphidlemachines.co.uk1 points·by smaddrellmander·2 tháng trước·0 comments
Softmax, can you derive the Jacobian? And should you care?idlemachines.co.uk134 points·by smaddrellmander·3 tháng trước·47 comments
Gemma 4 is not your standard transformeridlemachines.co.uk2 points·by smaddrellmander·3 tháng trước·0 comments
smaddrellmander·4 ngày trước·discussLove this! Cute graphics and the way the balls combine is really fun. Great job.
smaddrellmander·2 tháng trước·discussA similar vein, asking why in PRs. Most people don’t do the above, and asking why when it doesn’t make sense really helps understanding what your colleagues are actually thinking. Helps create a unified perspective on the codebase.