Fine-tune an 8B model on a 4 GB laptop GPU. Layer streaming keeps the frozen base out of VRAM and feeds it to the GPU one decoder layer at a time. 32 GB.
Global
> 第八章聚焦 GRPO、OPD、Search-R1 与 ReTool。更多 Agentic RL 算法、训练代码与实验实践，可以前往作者持续维护的另一个仓库 agentic-rl-lab；该仓库更新频率更高，会持续跟进新的算法与环境。
#1517 · +20 · 1d ·GitHub
🐧 PenguinHarness: Local-first multi-agent app development platform - create, fine-tune, deploy AI apps automatically!
🍔 In addition, ms-swift integrates the latest training technologies, including Megatron parallelism techniques such as TP, PP, CP, EP to accelerate training, as well as numerous GRPO algorithm family reinforcement learning algorithms including: GRPO, DAPO, GSPO, SAPO, CISPO…
#2760 · +8 · 1d ·GitHub
本项目旨在分享大模型相关技术原理以及实战经验（大模型工程化、大模型应用落地）
Fine-tuning large pretrained models is often prohibitively costly due to their scale.
Halo is an open-source framework built by White Circle for training large language and multimodal models
QLoRA uses bitsandbytes for quantization and is integrated with Hugging Face's PEFT and transformers libraries. QLoRA was developed by members of the University of Washington's UW NLP group.