métier
non classé
description
Evidence-based model-training recipes and run diagnosis (煉丹調參). Use for pretraining, midtraining, SFT, preference tuning, RLVR, LoRA/QLoRA, or speech training; training hyperparameters, loss spikes, data curation, compute budgets, stage evaluation, and…
Langue du texte source : anglais
mis à jour