https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's tech blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, distributed training, and data engineering. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Sat, 16 May 2026 00:00:00 GMT https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive.en A complete map of the 2026 LLM fine-tuning ecosystem. Open-source frameworks like Axolotl, Unsloth, LLaMA-Factory, TRL, PEFT, and TorchTune. LLM Foundry (MosaicML, acquired by Databricks). Cloud fine-tuning APIs from Modal, Together, OpenAI, Anthropic, and Cohere. Distributed training techniques like QLoRA, FSDP, and DeepSpeed Zero. Preference-optimization algorithms like DPO, GRPO (DeepSeek R1), KTO (Kahneman-Tversky), and IPO. Plus case studies from Korea (Upstage, KT, LG AI) and Japan (Sakana, Stockmark, ELYZA, PFN). Includes a decision guide for solo developers, academic researchers, startups, and enterprises. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) llmfinetuningaxolotlunslothllama-factorytrlpefttorchtunemosaicmlllm-foundrymodaldpogrpoktoqlorafsdpdeepspeed2026deep-diveenglish https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive.ja 2026年のLLMファインチューニング生態系を一気に整理する。Axolotl・Unsloth・LLaMA-Factory・TRL・PEFT・TorchTuneといったオープンソースフレームワークから、LLM Foundry(MosaicML、Databricksが買収)、Modal・Together・OpenAI・Anthropic・Cohereのクラウドファインチューニング API まで。QLoRA・FSDP・DeepSpeed Zero などの分散学習手法、DPO・GRPO(DeepSeek R1)・KTO(Kahneman-Tversky)・IPO といった選好最適化アルゴリズム、そして韓国(Upstage・KT・LG AI)・日本(Sakana・Stockmark・ELYZA・PFN)の事例まで。個人開発者・学術研究者・スタートアップ・エンタープライズそれぞれが何を選べば良いかの意思決定ガイドも収録。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) llmfinetuningaxolotlunslothllama-factorytrlpefttorchtunemosaicmlllm-foundrymodaldpogrpoktoqlorafsdpdeepspeed2026deep-dive日本語 https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-llm-finetuning-frameworks-2026-axolotl-unsloth-llama-factory-trl-peft-torchtune-deep-dive 2026년 LLM 파인튜닝 생태계를 한 번에 정리한다. Axolotl·Unsloth·LLaMA-Factory·TRL·PEFT·TorchTune 같은 오픈소스 프레임워크부터 LLM Foundry(MosaicML, Databricks 인수), Modal·Together·OpenAI·Anthropic·Cohere의 클라우드 파인튜닝 API까지. QLoRA·FSDP·DeepSpeed Zero 같은 분산 학습 기법, DPO·GRPO(DeepSeek R1)·KTO(Kahneman-Tversky)·IPO 같은 선호 최적화 알고리즘, 그리고 한국(Upstage·KT·LG AI)·일본(Sakana·Stockmark·ELYZA·PFN)의 사례까지. 1인 개발자·학술 연구자·스타트업·엔터프라이즈 각각이 무엇을 골라야 하는지 결정 가이드도 포함한다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) llmfinetuningaxolotlunslothllama-factorytrlpefttorchtunemosaicmlllm-foundrymodaldpogrpoktoqlorafsdpdeepspeed2026deep-dive