-
refactor: consolidate pure 32L Dense architecture, purge MoE files, audit hardening
-
Audit fix: Update tokenizer/configs/tokenizer_config.json for ViuMini-Dense-360M
-
Audit fix: Update tokenizer/configs/tokenizer_config.yaml for ViuMini-Dense-360M
-
fix: DDP auto_tune probe-before-wrap + broadcast micro/accum, R0-only logs, empty-shard guard, safety import path
-
remove v1 tokenizer (stale, mask typo) — use v2 from dataset repo tokenizer/tokenizer.json
-
sync local: enforce_mix sampler, tokenizer v2 (mask fix + 35/30/35), mix optional docs, repair scripts
-
FINAL 48k tokenizer (47756, unk0)
kaggle
-
docs: rename + push log
-
init: 28L MoE 241M audit-fixed, smoke pass