All repos — GH Trending Explorer

AirLLM 2.8T LLM inference with single 4GB GPU

Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)

Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.

Accessible large language models via k-bit quantization for PyTorch.

Halo is an open-source framework built by White Circle for training large language and multimodal models

Single-GPU LLM fine-tuning preflight: memory estimates, runnable LoRA/QLoRA recipes, and local measurements.

verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.

Open-source agent execution-recovery controller for coding and tool-using AI agents.

Open Source Instruction Tuning LLM Training Library 2026