Skip to main content
Ctrl+K

XTuner 0.2.0 documentation

Getting Started

  • Installation
  • Language Model Fine-tuning
  • Multimodal Large Model Fine-tuning
  • [Beta] RL: GRPO Training GSM8K

Pretraining & Fine-tuning

  • Fine-tuning Large Models with Trainer
  • Fine-tuning Multimodal Large Models with Trainer
  • Training Configuration
  • Dataset
  • Chat Template Description

Reinforcement Learning

  • Reinforcement Learning
    • [Beta] Customizing GRPO Training with Python Code

Advanced Tutorial

  • Fine-tuning & Pretraining
    • Model
    • Custom Dataset
    • Loss Function
    • FP8 Training
    • Performance Analysis
  • Reinforcement Learning
    • Model
    • Custom Dataset
    • Advanced Usage of RL Trainer
    • Loss Function

Benchmark

  • Megatron MoE Training Benchmark and Tuning Guide

Legacy Documentation

  • Welcome to XTuner Chinese Documentation

API

  • Pretrain & SFT Trainer
    • xtuner.v1.train.trainer.Trainer
    • xtuner.v1.train.toy_tokenizer.UTF8ByteTokenizer
  • Config
    • xtuner.v1.config.FSDPConfig
    • xtuner.v1.config.OptimConfig
    • xtuner.v1.config.AdamWConfig
    • xtuner.v1.config.LRConfig
    • xtuner.v1.config.GenerateConfig
  • RL Trainer
    • xtuner.v1.train.rl_trainer.RLColocateTrainer
    • xtuner.v1.train.rl_trainer.RLColocateTrainerConfig
    • xtuner.v1.train.rl_trainer.RLDisaggregatedTrainer
    • xtuner.v1.train.rl_trainer.RLDisaggregatedTrainerConfig
  • RL Config
    • xtuner.v1.rl.utils.AcceleratorResourcesConfig
    • xtuner.v1.rl.utils.CPUResourcesConfig
    • xtuner.v1.rl.rollout.worker.RolloutConfig
    • xtuner.v1.rl.agent_loop.SingleTurnAgentLoopConfig
    • xtuner.v1.rl.agent_loop_manager.AgentLoopManagerConfig
    • xtuner.v1.rl.agent_loop_manager.TaskSpecConfig
    • xtuner.v1.rl.agent_loop_manager.SamplerConfig
    • xtuner.v1.rl.agent_loop_manager.SyncProduceStrategyConfig
    • xtuner.v1.rl.agent_loop_manager.AsyncProduceStrategyConfig
    • xtuner.v1.rl.judger.JudgerConfig
    • xtuner.v1.rl.judger.GSM8KJudgerConfig
    • xtuner.v1.rl.judger.ComposedJudgerConfig
    • xtuner.v1.rl.replay_buffer.SyncReplayBufferConfig
    • xtuner.v1.rl.replay_buffer.AsyncReplayBufferConfig
    • xtuner.v1.rl.evaluator.EvaluatorConfig
    • xtuner.v1.rl.trainer.WorkerConfig
    • xtuner.v1.rl.loss.BaseRLLossConfig
    • xtuner.v1.rl.loss.GRPOLossConfig
    • xtuner.v1.rl.loss.OrealLossConfig
    • xtuner.v1.rl.rollout_is.RolloutImportanceSampling
  • Loss Context
    • xtuner.v1.loss.ce_loss.CELossConfig
    • xtuner.v1.loss.ce_loss.CELossKwargs
    • xtuner.v1.loss.ce_loss.CELossContext
    • xtuner.v1.rl.loss.BaseRLLossConfig
    • xtuner.v1.rl.loss.BaseRLLossKwargs
    • xtuner.v1.rl.loss.BaseRLLossContext
    • xtuner.v1.rl.loss.GRPOLossConfig
    • xtuner.v1.rl.loss.GRPOLossKwargs
    • xtuner.v1.rl.loss.GRPOLossContext
    • xtuner.v1.rl.loss.OrealLossConfig
    • xtuner.v1.rl.loss.OrealLossKwargs
    • xtuner.v1.rl.loss.OrealLossContext
  • .rst

Benchmark

  • Megatron MoE Training Benchmark and Tuning Guide

previous

Loss Function

next

Megatron MoE Training Benchmark and Tuning Guide

By XTuner Contributors

© Copyright 2024, XTuner Contributors.