-
Notifications
You must be signed in to change notification settings - Fork 149
Pull requests: thinking-machines-lab/tinker-cookbook
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
supervised/train.py: shift evals so they occur after 0, k, 2k, ... steps instead of 1, k+1, 2k+1, etc where k=config.eval_every
#90
opened Nov 12, 2025 by
joschu
Loading…
Add configurable temperature parameter for RL rollout sampling
#86
opened Nov 11, 2025 by
Xiuyu-Li
Loading…
make deepseekv3 renderer work with system messages, add renderer that forces thinking
#79
opened Nov 10, 2025 by
joschu
Loading…
add support for model_path to verifiers evaluate script
#64
opened Oct 31, 2025 by
tmabraham
Loading…
ProTip!
no:milestone will show everything without a milestone.