HomeToolsMCPHow It WorksStoriesPhilosophyArchitectureStar on GitHub
All Tools
U
Fine-tuningFreemiumOpen Source

UNSLOTH

Train and run LLMs locally 30x faster with 70% less memory

Apache-2.0

ABOUT

Traditional LLM fine-tuning is slow, memory-intensive, and requires complex setup. Unsloth solves this by providing optimized custom kernels that make training up to 30x faster with 70-90% less memory, enabling developers to fine-tune large models on consumer hardware without sacrificing accuracy.

INSTALL
pip install unsloth

INTEGRATION GUIDE

1. Fine-tuning LLMs like Mistral, Gemma, Llama, and Qwen for custom tasks 2. Running models locally with an OpenAI-compatible API and tool-calling support 3. Creating custom training datasets from PDFs, CSVs, and JSON files 4. Fine-tuning vision, audio, and embedding models efficiently 5. Exporting fine-tuned models to GGUF, Safetensors, Ollama, or vLLM formats

TAGS

llmfine-tuninginferencelocal-aiopen-sourceloragguftrainingaimachine-learning