Tinkerapi Seeks Post-Training Hackers To Advance Model Fine-Tuning

@johnschulman2 Very exciting!! Congrats on the success so far

@johnschulman2 dmed :))

@johnschulman2 lets go!!!

@johnschulman2 @miramurati Aye

Inspired by Horace He's @thinkymachines post on LLM Nondeterminism, we are building RL-Kernel for scalable post-training.

Single-GPU batch-invariant ops are nearly locked. Now tackling the real beast: cross-TP determinism (vLLM TP>1 vs FSDP TP=1) & the memory wall.

Would love your harsh feedback on our roadmap! https://github.com/RL-Align/RL-Kernel/issues/83

@johnschulman2 @miramurati 👀

@johnschulman2 @miramurati Hackers the word is so hyper and underrated!

@johnschulman2 will tinker make bad fine-tune runs inspectable since most teams fail there

@johnschulman2 I have some good red teaming experience on frontier models, Dmed!

@johnschulman2 Tinker is so much underrated IMO, would love to help make it better😄 (it's a bit too expensive for part-time researcher without grant tho....