Miles v0.1.2 adds experimental Kubernetes backend and torchtitan training backend

RadixArk has released Miles v0.1.2, adding an experimental native Kubernetes backend so RL jobs can run as ordinary cluster workloads, and torchtitan as a third training backend alongside Megatron and FSDP. The announcement, posted on X by RadixArk, says orchestration can restart while training continues.

The release also adds score centering for stable async RL and support for DeepSeek-V4.1-Flash and MiMo-V2.6-Flash-RL. The Kubernetes backend is described as experimental, and the torchtitan backend is presented as another PyTorch-native training option.