Hugging Face's logo
Konstantin's picture

Konstantin PRO

apsys
whitecircle
39 2 11
Kirili4ik's profile picture dm33tri's profile picture GigaBoy's profile picture
Ā·

AI & ML interests

Interpretable ML, classical ML, LLM & inference opts

Recent Activity

reacted to medmekk's post with šŸš€ 2 days ago
šŸš€ Introducing Halo 1.0 Today, we are open-sourcing Halo, the training framework we use to train every model at White Circle. It comes with: šŸ§  Full post-training stack: SFT, DPO/KTO/SMPO, reward modeling, GRPO, distillation šŸ¤– Async multi-turn RL with vLLM/SGLang rollouts and sandboxed tool use āš” ~2.8Ɨ TRL throughput on 8Ɨ B300 (EP+FSDPv2, FA4, fp8/fp4) šŸ¤— Dense HF models + 15 MoE families (Qwen, GLM, Mistral, DeepSeek-V4ā€¦) šŸ› ļø One halo command, prebuilt Docker images, and docs for humans and agents šŸ’» https://github.com/whitecircle/halo Try it and tell us what you're training
liked a model 8 days ago
whitecircle/GLM-4.7-Flash-Coder
posted an update 14 days ago
43% less answer time in a looped transformer by changing answer supervision check out https://x.com/advprop/status/2098087083470373010?s=20 + full writeup linked there
View all activity

Organizations

Blog-explorers's profile picture vcsurf's profile picture Vikhr models's profile picture GEN-Enterprise's profile picture White Circle's profile picture