post-training-slime

原文涵盖的主题

• slime: LLM Post-Training Framework for RL Scaling
• When to Use slime
• Key Features
• Architecture Overview
• Installation
• From Source
• Quick Start: GRPO Training
• Workflow 1: Standard GRPO Training

slime is an LLM post-training framework from Tsinghua's THUDM team, powering GLM-4.5, GLM-4.6, and GLM-4.7. It connects Megatron-LM for training with SGLang for high-throughput rollout generation. Choose slime when you need: Megatron-LM native training with SGLang inference Custom data generation workflows with flexible data buffers Training GLM, Qwen3, DeepSeek V3, or Llama 3 models Research-grade framework with production backing (Z.ai) Consider alternatives when: You need enterprise-grade stability features → use miles You want flexible backend swapping → use verl You need PyTorch-native abstractions → use torchforge Training: Megatron-LM with full parallelism support (TP, PP, DP, SP) Rollout: SGLang-based high-throughput generation with router Data Buffer: Flexible prompt management and sample storage Models: GLM-4.x, Qwen3, DeepSeek V3/R1, Llama 3 -- Use this workflow for training reasoning models with group-relative advantages. [ ] Docker environment or Megatron-LM + SGLang installed [ ] Model checkpoint (HuggingFace or Megatron format) [ ] Training data in JSONL format Or with chat format: Choose a pre-configured model script: [ ] Check TensorBoard: tensorboard --logdir outp

这里只展示节选，请以完整原文和当前要求为准。