d37bfa9d75
[update] harden training and inference reliability
...
Fix truncated SFT/DPO label masking and resume-safe gradient accumulation to prevent unstable optimization, and add explicit eval backend selection plus regression checks for safer local validation.
2026-04-05 18:11:37 +00:00
101d7df2da
[update] minimind-3
2026-03-25 23:57:45 +08:00
349e74ec7b
[update] empty_think_ratio
2026-02-06 19:15:21 +08:00
288e1ac02a
[update] empty_think_ratio
2026-02-06 01:36:02 +08:00
ccc190da05
[feat] data process
2026-02-06 01:17:57 +08:00
3a5aba82db
[fix] max length
2026-01-17 13:26:14 +08:00
714abcf802
[update] pretrain load
2026-01-17 12:00:17 +08:00
aa539a824a
[update] align mask
2026-01-15 11:20:41 +08:00
c090b69c4d
[update] align loss
2026-01-15 00:56:32 +08:00
f55d4c32a0
[update] mask log
2026-01-07 22:12:26 +08:00
7641985d14
[update] simplify loader
2026-01-06 01:20:52 +08:00
4a5c9f5ece
[feat] stream load data
2025-12-28 16:58:52 +08:00
fe24501602
[feat] adjust seq length
2025-12-14 20:41:58 +08:00
ce693f8e7f
[fix] git track
2025-10-21 23:09:00 +08:00
7d877db79b
[feat] minimind-2510
2025-10-21 21:19:47 +08:00
274483cb1b
250426
2025-04-26 10:07:55 +08:00
root
jingyaogong