Tinker
@tinkerapi
I tink, therefore I am.
Post-training API by @thinkymachines
加入 January 2026
1 正在关注 13.4K 粉丝
Our friends + occasional antagonists at @SemiAnalysis_ published a great writeup on RL training efficiency: treat the system as a queue and keep generator and trainer throughput matched.
Also includes an analysis of Tinker's cost-efficiency and many OSS RL frameworks!
显示更多
RL Systems Mind the Gap:
Matching Trainer and Generator Throughput
RL Training Infrastructure, GRPO,
PipelineRL, Async RL, Policy Staleness,
RL Sandbox Infra, CPU Requirements,
TCO Analysis, Thinking Machines Tinker
显示更多
