unsloth

Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.

View on GitHub↗Upstream unslothai/unsloth↗

Builder

Unsloth

unslothai • startup

Stars

59,074

Using upstream star count

Forks

5,010

Using upstream fork count

Open Issues

Activity Score

0/100

0 commits in 30d

Created

Nov 29, 2023

Project creation date

README Summary

Unsloth is a Python library that optimizes fine-tuning and reinforcement learning for large language models including GPT, DeepSeek, Qwen, Llama, Gemma, and TTS models. It provides 2x faster training speeds while using 70% less VRAM compared to standard training methods. The library focuses on memory efficiency and performance optimization for LLM training workflows.

AI Dev Skills

Unmapped

Tags

LoRA Fine-tuningQLoRA Fine-tuningReinforcement Learning from Human FeedbackMemory OptimizationCUDA ProgrammingTransformer ArchitectureGradient CheckpointingMixed Precision TrainingParameter-Efficient Fine-tuningModel QuantizationDistributed TrainingTextOpen Source LLMsCodeMemory-Efficient TrainingLarge Language Model TrainingLanguage Model Fine-tuningAudioGPU ClustersParameter-Efficient TrainingCloud TrainingText-to-Speech Model TrainingSelf-hostedConversational AI DevelopmentQLoRA OptimizationGradient OptimizationDemocratized AI TrainingLow-Resource TrainingTransformer Architecture OptimizationCUDA Kernel DevelopmentMulti-language Model AdaptationCustom Model DevelopmentInstruction Following TrainingMemory-Optimized AICode Generation Model TrainingColab/KaggleCustom Domain AdaptationPython

Recent Activity

Updated 1 months ago

7 Days

30 Days

90 Days

Quality

production

Quality: high
Maturity: production

PM Skills

Developer Platform

Languages

Python100.0%

Timeline

Project created: Nov 29, 2023
Forked: Mar 12, 2026
Your last push: 1 months ago
Upstream last push: 6 days ago
Tracked since: Mar 12, 2026

Similar Repos

pgvector cosine similarity · $0

Loading…