llm-training

SKILL.md

Frameworks and techniques for training and finetuning large language models.

| Framework | Best For | Multi-GPU | Memory Efficient |

| Accelerate | Simple distributed | Yes | Basic | | DeepSpeed | Large models, ZeRO | Yes | Excellent | | PyTorch Lightning | Clean training loops | Yes | Good | | Ray Train | Scalable, multi-node | Yes | Good | | TRL | RLHF, reward modeling | Yes | Good | | Unsloth | Fast LoRA finetuning | Limited | Excellent |

Use when "training LLM", "finetuning", "RLHF", "distributed training", "DeepSpeed", "Accelerate", "PyTorch Lightning", "Ray Train", "TRL", "Unsloth", "LoRA training", "flash attention", "gradient checkpointing" Source: eyadsibai/ltk.

View raw