技能低风险未认领

TRL Training on Hugging Face Jobs

Train or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, model selection/leaderboards and model persistence. Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.

huggingfacehuggingface/huggingface-llm-trainer★ 11k更新于 2026年10月1日

说明

它的许可证不允许 Harness 保存副本,说明请到来源查看。huggingface/skills ↗

权限

声明检测
运行代码—python
安装—brewpipsystem packagesuvx
安装时运行脚本—无
网络—datasets-server.huggingface.codiscuss.huggingface.codocs.astral.shgist.githubusercontent.comgithub.comhf.cohuggingface.colmstudio.aiollama.airaw.githubusercontent.comunsloth.ai
需要的凭据—HF_TOKEN
工作区外的路径—无
智能体工具—无

检查

低风险 · 没有发现需要提醒的地方。

未经人工审核 · 已做规则检查;模型审核尚未开启。

另有 3 处低风险标记:常见命令之类,只记录、不提醒
  • 规则 · privilege_escalationscripts/convert_to_gguf.py:27
  • 规则 · privilege_escalationscripts/convert_to_gguf.py:28
  • 规则 · privilege_escalationscripts/convert_to_gguf.py:57

文件19 个文件 · 182.1 KB

  • SKILL.md28.1 KB
references/10
  • gguf_conversion.md9.6 KB
  • hardware_guide.md6.6 KB
  • hub_saving.md8.3 KB
  • local_training_macos.md8.1 KB
  • reliability_principles.md10.6 KB
  • trackio_guide.md6.3 KB
  • training_methods.md4.9 KB
  • training_patterns.md6.0 KB
  • troubleshooting.md8.7 KB
  • unsloth.md7.8 KB
scripts/8
  • convert_to_gguf.pyx12.3 KB
  • dataset_inspector.py15.3 KB
  • estimate_cost.pyx4.7 KB
  • hf_benchmarks.pyx19.6 KB
  • train_dpo_example.py3.0 KB
  • train_grpo_example.py2.3 KB
  • train_sft_example.py3.3 KB
  • unsloth_sft_example.py16.5 KB

它的许可证不允许 Harness 分发副本:Codeg 按这个提交从 GitHub 取文件,并核对哈希。

版本

  1. #1—最新2026年10月7日