Will LinandClaude Fable 5 37aa382cce [bugfix] example: the distilled FastH3 grid is 4 forwards = a 5-point sigma grid ...

MiniMaxH3Scheduler.set_timesteps(N) builds an N-point sigma grid ending at
0 and runs N-1 transformer forwards (the base model's '50 steps' preset is
49 forwards). The student was distilled on a 4-FORWARD grid
(t = 1000/750/500/250 -> 0 on the shift-12 schedule), so --steps 4 was
silently running a 3-forward, off-distribution grid. Default is now 5
grid points = the distilled 4-forward grid, with the convention documented
on the flag.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-08-21 09:27:38 +00:00

..

datasets

[misc]: consolidate dataset download scripts under examples/datasets/ (#1667)

2026-08-02 14:21:24 -07:00

distill

[misc]: consolidate dataset download scripts under examples/datasets/ (#1667)

2026-08-02 14:21:24 -07:00

inference

[bugfix] example: the distilled FastH3 grid is 4 forwards = a 5-point sigma grid

2026-08-21 09:27:38 +00:00

serving

[infra] Dreamverse 09/14: Add Docker and launch scripts (#1332)

2026-05-12 18:01:11 -07:00

train

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

training

[bugfix] profiler region system: record + export actually work, usability roll-up (#1691)

2026-08-09 20:35:53 -07:00