Write the reference before debugging the shader
The Flash Attention 2 path was easier to finish after the algorithm existed in plain JavaScript...
Tag archive
The Flash Attention 2 path was easier to finish after the algorithm existed in plain JavaScript...
The WebGPU path did not have one honest speedup number. Against multithreaded WASM SIMD, the...
The WebGPU matmul sweep started with a naive kernel, then added 16 by 16 workgroup tiling and a 4 by...
Over the past few days, MyZubster has evolved from a collection of ideas, repositories, and visual...
A 32-bit WebAssembly heap put the browser playground near a 4 GB ceiling. With weights and AdamW...
PostTrainLLM's curriculum begins with a model as a parameterized function and ends with evaluation...
시가총액 데이터가 전부 비어버리자 종목 구성 순위가 알파벳순으로 둔갑했고, 이를 막는 3중 방어층을 새로 넣었습니다. 시가총액이 전부 비어서 생긴 일 오늘 아침...
When market-cap data came back entirely empty, stock rankings silently reverted to alphabetical order...
새 공유계좌로 실거래를 시작했고, 그 과정에서 계좌 구분값을 안 넘기거나 전역으로 공유해버리는 같은 유형의 버그가 킬스위치 오발과 강제매도로 여러 번 반복됐습니다 ...
I started live trading through a new shared account, and along the way the same class of bug kept...
Yesterday I patched the symptom of a false kill-switch trip in a hurry. Today's postmortem found the...
One early fine-tune appeared to beat its baseline. The result disappeared when the baseline and...