mojolearn: GPU machine-learning algorithms in Mojo, portable to Apple Silicon

Le résumé fourni par la source

Ports of GPU machine-learning algorithms into Mojo from the CUDA implementations that established them, targeting Apple Silicon via Metal first and CUDA and ROCm from the same source. Gradient-boosted oblivious (symmetric) trees, random forests and extremely randomized trees mirror CatBoost; k-means and brute-force k-nearest-neighbors mirror cuVS; DBSCAN and PCA mirror cuML; ordinary least squares and top-k selection mirror RAFT. Every derivation is recorded per file in PORTED_MAP.tsv and the license obligations each carries are in NOTICE. Each estimator runs in a FAST mode and an IDENTICAL mode whose results are bit-identical across Metal, HIP and CUDA; the cross-vendor certificates are in E1_RESULTS.md and E2_RESULTS.md.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.