​

Help System

​

General Help

aitraining --help

• Available subcommands
• Global options
• Usage examples

​

Command Help

aitraining llm --help
aitraining text-classification --help
aitraining chat --help

​

Trainer-Specific Help

# SFT trainer parameters
aitraining llm --trainer sft --help

# DPO trainer parameters
aitraining llm --trainer dpo --help

# ORPO trainer parameters
aitraining llm --trainer orpo --help

# PPO trainer parameters
aitraining llm --trainer ppo --help

​

Preview Mode

aitraining llm --preview-trainer dpo --help

​

Parameter Groups

• Basic - Model, data, project settings
• Data Processing - Text columns, chat templates, tokenization
• Training Configuration - Trainer, logging, saving
• Training Hyperparameters - LR, epochs, batch size
• PEFT/LoRA - Adapter settings
• DPO/ORPO - Preference learning params
• Hub Integration - Push to Hugging Face
• Knowledge Distillation - Teacher-student params
• Hyperparameter Sweep - Auto-tuning options
• Enhanced Evaluation - Advanced metrics
• Advanced Features - Custom loss, custom metrics
• Reinforcement Learning - PPO settings
• Inference - Inference prompts and generation params

​

Finding Parameters

​

By Name

aitraining llm --help | grep lora

​

By Group

PEFT/LoRA:
  --peft                Enable LoRA training
  --lora-r              LoRA rank (default: 16)
  --lora-alpha          LoRA alpha (default: 32)

​

Quick Reference

​

Next Steps