Introducing Halo, the best framework for post-training of open-source models.
@whitecircle2.5K followers
2,343,658 views
56 Comments
Comments from the original post on X (Twitter)

@whitecircle
Every model we train at White Circle now uses Halo. The same codebase runs LoRA on a 24 GB GPU, multi-node training on B300s, and async RL. Technical blog:
40 likes

@whitecircle
Other training frameworks often require a separate implementation for each model family. In Halo, adding a new model family takes just 100 lines of connecting a wrapper instead of rewriting the model.
33 likes

@whitecircle
One YAML file contains the model, training method, GPU setup and checkpoint settings. One command starts training.
29 likes

@whitecircle
Halo supports data, tensor, context, expert and expert-tensor parallelism. It also includes optimized attention, grouped-GEMM and fused loss kernels.
29 likes

@whitecircle
Halo is not limited to supervised fine-tuning. It supports async RL and training with external environments. We are excited to partner with the @sgl_project team to make it the primary engine for rollouts.
37 likes

@whitecircle
On @OpenAI gpt-oss-20b, Halo delivered 2.3–2.8x the throughput of stock TRL with less peak memory. Both runs used the same FlashAttention, Liger, fused cross-entropy and grouped-GEMM optimizations.
30 likes

@whitecircle
We partnered with @liquidai to add native Halo support for LFM2.5-8B-A1B. On one B300, Halo was up to 20% faster than next best open-source framework.
35 likes

@whitecircle
We also used Halo to fine-tune @Zai_org GLM-4.7-Flash on 177M tokens of agentic traces. The resulting model improved SWE-rebench-V2 by @nebiusai from 33% to 42%. Halo reached up to 1.63× TRL throughput on the same model precision and data. Model and write-up:
30 likes

@whitecircle
Thanks for reading to the end! Star us on GitHub: Read more:
35 likes

@DanielSmidstrup
Open source is the way, and it's only going to get more and more relevant in the coming time! :)
10 likes

@whitecircle
True
4 likes

@leo1yu
insane way to kick off the week -- congrats @whitecircle team
9 likes

@whitecircle
Lets go!
3 likes

@cneuralnetwork
crazy reading through the docs
8 likes

@vishalsingh2972
love that this is open source, congrats team ❤️
8 likes

@catalinmpit
Congrats on the awesome work, team! Good to see training infra being open source 👏 Starred the repo to contribute to the growth.
8 likes

@whitecircle
thanks!
2 likes

@pmitu
great idea you've got my star, guys
8 likes

@giutheginger
took me one read to understand exactly who this is for. that almost never happens with infra launches. congrats team!!! starred 👀
7 likes

@MacdevM
Peak improvement, I will try it from first hand
7 likes

























