White Circle 推出 Halo 框架：开源模型后训练吞吐量达 TRL 的 2.8 倍
想自己微调模型可以看看 White Circle 的 Halo，吞吐量达 TRL 的 2.8 倍、更省内存，还不用转格式，已开源。
White Circle 发布开源框架 Halo，定位为开源模型后训练工具。官方数据显示其吞吐量最高达到原版 TRL 的 2.
8 倍，峰值内存占用更低。Halo 让模型保持原生 HuggingFace 格式，训练后无需额外转换。
代码已在 GitHub 开源，Hugging Face CEO Clement Delangue 转发推荐，称训练模型正变得越来越容易。
Training models is becoming easier and easier - just look at this and TRL - especially with agents! You're missing out if you're still using off the shelf models for all your tasks! White Circle @whitecircle Introducing Halo, the best framework for post-training of open-source models. Halo delivers up to 2.8x the throughput of stock TRL with less peak memory, while models stay in their native HuggingFace format. Star us on GitHub: github.com/whitecircle/ha…
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Halo, TRL, HuggingFace, 开源模型.
Source attribution: curated from Clement Delangue (@ClementDelangue).
