skills.rest
Junkfooooood
JunkfoooooodCOMMUNITY
Explore all skills by this author →
GitHub

axolotl

Configure and run LLM fine-tuning with Axolotl YAML configs for LoRA, DPO, and GRPO.

Updated May 4, 2026
Search Tags:#distributed training#lora#dpo#grpo#axolotl#llm fine-tuning#yaml config
One-click install
npx skills add https://github.com/Junkfooooood/hermes-governance --skill axolotl-junkfooooood
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: axolotl
Source: https://github.com/Junkfooooood/hermes-governance/tree/main/skills/mlops/training/axolotl
Command: npx skills add https://github.com/Junkfooooood/hermes-governance --skill axolotl-junkfooooood

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires axolotl, torch, transformers, datasets, peft, accelerate, deepspeed, and includes references (resource) components.

What problem does it solve? Fine-tuning large language models requires coordinating datasets, tokenization, distributed training, and RLHF methods, which is error-prone when done by hand. This Skill provides expert guidance for Axolotl, a YAML-driven framework that handles LoRA/QLoRA, DPO, KTO, ORPO, GRPO, and multimodal training across 100+ models. ## Core Features & Use Cases - YAML-based fine-tuning: Configure SFT, pre-training, and preference-based post-training (DPO, KTO, ORPO, GRPO) through declarative config files. - Distributed training support: Set up FSDP, DeepSpeed, context parallelism, and mixed precision (BF16/FP8) for multi-GPU workloads. - Dataset format guidance: Work with conversation, instruction, template-free, and pre-tokenized dataset formats with chat_template support. - Use Case: Fine-tune a Llama model with QLoRA on a custom chat dataset by generating the correct Axolotl YAML config, validating dataset formatting, and debugging NCCL or FSDP issues. ## Quick Start Ask the assistant to generate an Axolotl YAML config for fine-tuning a specific model with LoRA on your dataset format.

Frequently Asked Questions about axolotl

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fine-tune an LLM with Axolotl?▼

Create a YAML config specifying base_model, datasets with their type, and adapter settings like lora, then run the axolotl train CLI command. Axolotl handles tokenization, dataset preprocessing, and training loop setup automatically.

What dataset formats does Axolotl support?▼

Axolotl supports pre-training text, conversation formats with chat_template, instruction formats like Alpaca, template-free input_output with segment masking, and pre-tokenized datasets. Preference methods like DPO use chosen/rejected pair formats.

Does Axolotl support multi-GPU and distributed training?▼

Yes, Axolotl supports FSDP, DeepSpeed, and context parallelism for multi-GPU training. Configure fsdp_config in the YAML with options like transformer_layer_cls_to_wrap, and ensure context_parallel_size divides your total GPU count.

How do I train with DPO or GRPO in Axolotl?▼

Set rl: dpo or configure GRPO in your YAML and provide preference datasets with chosen/rejected pairs using types like chatml.intel. GRPO uses vLLM for trajectory generation and supports custom reward functions.

Why does Axolotl training hang or fail with NCCL errors?▼

Hangs usually indicate GPU communication issues between nodes. Run NCCL tests like all_reduce_perf to verify bandwidth, check that context_parallel_size divides your GPU count, and review the FAQ for deepspeed and padding token errors.

Can Axolotl fine-tune multimodal vision language models?▼

Yes, Axolotl has beta support for multimodal models including Llama 3.2 Vision, Llama 4, Pixtral, and Gemma 3. Set processor_type: AutoProcessor, disable sample_packing, and use the appropriate chat_template for your model.

Related Skills

Explore More Software Engineering Skills→
thinking-machines-labthinking-machines-lab
4.0k

datasets

Construct and normalize ML datasets across JSONL, HuggingFace, and conversation formats.

Official
Intermediate
kaminocorpkaminocorp
1

peft-fine-tuning

Fine-tune LLMs with LoRA, QLoRA, and adapter methods using HuggingFace PEFT.

Official
Advanced
twjarviszyra-webtwjarviszyra-web
3

evaluating-llms-harness

Evaluate LLMs against academic benchmarks with reproducible metrics.

Community
Advanced
xu1713xu1713

unsloth

Fine-tune LLMs with LoRA and QLoRA using Unsloth's optimized training framework.

Community
Intermediate