{
 "schema_version": 1,
 "generated": "2026-10-20",
 "repo": "whitecircle/halo",
 "name": "halo",
 "owner": "whitecircle",
 "slug": "whitecircle-halo",
 "github": "https://github.com/whitecircle/halo",
 "what": "Halo is an open-source framework from White Circle for training large language and multimodal models with pre-training, SFT, preference optimization, and multi-turn RL on Hugging Face models.",
 "why": "",
 "language": "Python",
 "requirements": [
  "Docker installed on a host with an NVIDIA GPU (Hopper H100/H200 or Blackwell B200/B300 images are published)",
  "Network access to public.ecr.aws to pull the Halo image anonymously",
  "A Hugging Face model or the example configs in the examples/ folder",
  "Enough GPU memory for the chosen model and method (the README shows single-GPU LoRA and 8-GPU runs)"
 ],
 "audience_note": "Written for ML engineers and AI agents who have access to NVIDIA Hopper or Blackwell GPUs and want to train or fine-tune Hugging Face models at scale.",
 "risk": "none",
 "section": "earlier",
 "weekly_rank": null,
 "category": "Software and DevOps",
 "url": "https://dedupedai.com/use-cases/whitecircle-halo/",
 "use_case_count": 5,
 "use_cases": [
  {
   "id": "whitecircle-halo--lora-sft-single-gpu",
   "title": "Fine-tune a Hugging Face LLM with LoRA on one GPU",
   "best": true,
   "url": "https://dedupedai.com/use-cases/whitecircle-halo/lora-sft-single-gpu/",
   "json": "https://dedupedai.com/use-cases/u/whitecircle-halo/lora-sft-single-gpu.json"
  },
  {
   "id": "whitecircle-halo--moe-expert-parallel-sft",
   "title": "Train a mixture-of-experts model across 8 GPUs",
   "best": false,
   "url": "https://dedupedai.com/use-cases/whitecircle-halo/moe-expert-parallel-sft/",
   "json": "https://dedupedai.com/use-cases/u/whitecircle-halo/moe-expert-parallel-sft.json"
  },
  {
   "id": "whitecircle-halo--async-rl-with-environments",
   "title": "Run multi-turn RL against vLLM or SGLang rollouts",
   "best": false,
   "url": "https://dedupedai.com/use-cases/whitecircle-halo/async-rl-with-environments/",
   "json": "https://dedupedai.com/use-cases/u/whitecircle-halo/async-rl-with-environments.json"
  },
  {
   "id": "whitecircle-halo--quantize-checkpoint-nvfp4",
   "title": "Quantize a trained checkpoint to NVFP4",
   "best": false,
   "url": "https://dedupedai.com/use-cases/whitecircle-halo/quantize-checkpoint-nvfp4/",
   "json": "https://dedupedai.com/use-cases/u/whitecircle-halo/quantize-checkpoint-nvfp4.json"
  },
  {
   "id": "whitecircle-halo--agent-driven-training-setup",
   "title": "Let a coding agent set up and launch training jobs",
   "best": false,
   "url": "https://dedupedai.com/use-cases/whitecircle-halo/agent-driven-training-setup/",
   "json": "https://dedupedai.com/use-cases/u/whitecircle-halo/agent-driven-training-setup.json"
  }
 ],
 "disclaimer": "Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk."
}