Report
Miles v0.1.1 adds Multi-LoRA and agent-training support
A post announcing the release says Tinker API-compatible Multi-LoRA lets multiple training jobs share a base model without interfering. It also says Miles can run Harbor tasks directly in sandboxes such as AgentENV, Daytona, E2B and Modal.
TLDR
A post announcing Miles v0.1.1 says the release works natively with agent-training harnesses like Claude Code and adds Tinker API-compatible Multi-LoRA. It claims larger-model training requires less memory, with validated support across NVIDIA and AMD GPUs. The post also lists Qwen3.8-Flash-Next, GLM-5.3-Flash and Kimi-K3 among the models now supported in the stable version.
Combined views
13
1 Source, first seen 3h ago
