The annotated PyTorch training loopidlemachines.co.uk81 points·by smaddrellmander·19 hari yang lalu·19 comments
Mae vs. MSE: more than just the mean vs. median debateidlemachines.co.uk1 points·by smaddrellmander·23 hari yang lalu·0 comments
DiffusionGemma: Discrete diffusion in a large language modelidlemachines.co.uk3 points·by smaddrellmander·bulan lalu·1 comments
Reading MAI's efficiency gain. How to pick architectures like serious peopleidlemachines.co.uk9 points·by smaddrellmander·bulan lalu·0 comments
MAI-Thinking-1: Building a Hill-Climbing Machine [pdf]microsoft.ai3 points·by smaddrellmander·bulan lalu·0 comments
Are contrastive losses just cross entropy all along?idlemachines.co.uk2 points·by smaddrellmander·bulan lalu·0 comments
Every token, everywhere, all at onceidlemachines.co.uk2 points·by smaddrellmander·2 bulan yang lalu·0 comments
The cut in the Mixture of Experts compute graphidlemachines.co.uk1 points·by smaddrellmander·2 bulan yang lalu·0 comments
DeepSeek V4 from the Insideidlemachines.co.uk2 points·by smaddrellmander·2 bulan yang lalu·0 comments
Softmax, can you derive the Jacobian? And should you care?idlemachines.co.uk134 points·by smaddrellmander·3 bulan yang lalu·47 comments
Gemma 4 is not your standard transformeridlemachines.co.uk2 points·by smaddrellmander·3 bulan yang lalu·0 comments
smaddrellmander·4 hari yang lalu·discussLove this! Cute graphics and the way the balls combine is really fun. Great job.
smaddrellmander·2 bulan yang lalu·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.