Axolotl: Flexible Multi-Method Training Framework

Quick Facts

Aspect	Details
Organization	Open Source Community
Purpose	Flexible, config-driven training
License	Apache 2.0
Installation	pip install axolotl
Best For	Flexible multi-method training
Interface	YAML configuration

What It Does

Axolotl is a user-friendly, highly-configurable training framework supporting: - LoRA and QLoRA - Full fine-tuning - Multiple training methods (SFT, ORPO, DPO) - Distributed training - Multiple data formats - Environment setup automation

Installation

pip install axolotl
# Or from source
git clone https://github.com/OpenAccess-AI-Collective/axolotl
cd axolotl
pip install -e .

Core Features

1. Configuration-Based Training

# config.yaml
base_model: meta-llama/Llama-2-7b-hf
model_type: llama
tokenizer_type: llama

# Data
dataset: alpaca
dataset_prepared_path: null
val_set_size: 0.1

# Quantization (optional)
load_in_8bit: false
load_in_4bit: false

# LoRA Configuration
lora_model_dir: null
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj
  - k_proj
  - o_proj

# Training
output_dir: ./qlora-output
num_epochs: 3
micro_batch_size: 4
gradient_accumulation_steps: 4
warmup_steps: 100
learning_rate: 2e-4
optimizer: paged_adamw_32bit

# Advanced
deepspeed: null
flash_attention: true

2. Running Training

# Train from config
axolotl train config.yaml

# Inference with trained model
axolotl inference config.yaml

# Merge LoRA adapters
axolotl merge_lora config.yaml \
  --lora_model_dir ./qlora-output \
  --output_dir ./merged-model

3. Multi-Method Support

# For DPO training
dataset: dpo_pairs
dataset_format: dpo

# For SFT
dataset_format: sharegpt

# For ORPO
dataset_format: conversation

Memory Profiles

# For consumer GPU (RTX 4090, 24GB)
load_in_4bit: true
bnb_4bit_quant_type: nf4
lora_r: 16
micro_batch_size: 4
gradient_accumulation_steps: 4

# For A100 (80GB)
load_in_8bit: false
lora_r: 32
micro_batch_size: 16
gradient_accumulation_steps: 2

Data Format Support

# Alpaca format
dataset_format: alpaca

# ShareGPT format (conversations)
dataset_format: sharegpt

# DPO format (preferred/rejected)
dataset_format: dpo

# Conversation format
dataset_format: conversation

Distributed Training

# For multi-GPU training
deepspeed: configs/deepspeed.json

# Model parallel (for very large models)
world_size: 4
rank: 0

Strengths

✅ Flexibility - Multiple training methods in one framework
✅ Configuration - No code needed, YAML-driven
✅ Production-Ready - Supports complex setups
✅ Data Formats - Handles various dataset formats
✅ Active Development - Regular updates

Weaknesses

❌ Speed - Slower than specialized tools like Unsloth
❌ Learning Curve - Many configuration options
❌ Memory - Similar to TRL without QLoRA benefits

Best Practices

1. Start Simple - Use basic config, then customize
2. Validate Data - Use the validation split
3. Monitor Training - Watch metrics during training
4. Use Flash Attention - Significantly speeds up training

Common Recipes

Recipe 1: Quick LoRA Fine-tuning

lora_r: 8
lora_alpha: 16
micro_batch_size: 8
gradient_accumulation_steps: 2
num_epochs: 1

Recipe 2: QLoRA on Consumer GPU

load_in_4bit: true
lora_r: 16
micro_batch_size: 4
gradient_accumulation_steps: 8
num_epochs: 3

Recipe 3: Full Training on A100

lora_model_dir: null  # No LoRA - full training
micro_batch_size: 32
gradient_accumulation_steps: 1
learning_rate: 1e-4

When to Use Axolotl

Scenario	Recommendation
Config-driven training	✅ Best
Multi-method experiments	✅ Best
Production training	✅ Good
Speed critical	❌ Use Unsloth
RLHF needed	⚠️ Use TRL

Resources

• GitHub
• Config Examples
• Discord Community

---

Next: Try 03 Unsloth for speed-optimized training