unsloth-zoo
Utils for Unsloth
Decision gist · record as of 2026-08-14
Yes, if you are fine-tuning large language models and want to reduce GPU memory usage and training time. The package is actively maintained with no known vulnerabilities and integrates with the standard Hugging Face ecosystem. The LGPL-3.0-or-later license is permissive for research and commercial use provided you distribute source code. Install friction is low. Verify that the specific models and hardware you plan to use are supported before committing to a large training run.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires PyTorch and CUDA-capable GPU or compatible accelerator; on Windows, PyTorch must be pre-installed before installing unsloth-zoo.
- Low install friction with a pure Python wheel.
- Active maintenance with recent releases and a large repository (71439 stars).
License · maintenance · safety
LGPL-3.0-or-later (copyleft) — Licensed under LGPL-3.0-or-later (copyleft). Derivative works and modifications must be distributed under compatible terms; commercial use is permitted if source code and license are provided.
last release 2026-08-14 (0 days) · last repo commit 2026-08-14 · 71,439 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 1,352,315 downloads/mo, #4,012 on PyPI
Alternatives
Verify before relying
pip install unsloth-zoo
from unsloth import FastLanguageModel
import torch
model, tokenizer = FastLanguageModel.from_pretrained(
model_name="unsloth/model",
load_in_4bit=True,
)
model = FastLanguageModel.get_peft_model(
model,
target_modules=["q_proj", "v_proj"],
bias="none",
use_gradient_checkpointing="unsloth",
)- Exact speedup and memory reduction percentages claimed in description for specific model/hardware combinations.
- Whether all listed model families (gpt-oss, Gemma 3n, Qwen3, Llama 4, Mistral) are fully supported in version 2026.8.12.
- Compatibility with non-NVIDIA GPUs beyond the Blackwell mention.
- Specific LoRA configuration parameters and default values supported by this version.
What it is and what it does
Unsloth Zoo is a utility package that provides optimized training recipes for large language models, integrating with PyTorch, transformers, and peft to enable memory-efficient fine-tuning. It applies techniques like quantization and gradient checkpointing to reduce GPU memory requirements and accelerate training. The package supports various model architectures and works with both consumer and enterprise GPUs.
Users typically load a pretrained model, apply optimization techniques, and train on custom datasets using the trl library for instruction tuning or reasoning tasks. The package includes support for distributed training through accelerate and can export fine-tuned models to various formats. It depends on 27 runtime packages including torch, transformers, trl, peft, and accelerate.
Use it for
- Fine-tune language models on consumer GPUs using quantization and memory-efficient kernels.
- Train reasoning models with reduced VRAM by combining gradient checkpointing and optimized operations.
- Adapt vision models for custom image-text tasks with reduced memory footprint.
- Export fine-tuned models to GGUF or Ollama format for local deployment after training.
- Implement text-to-speech model fine-tuning using memory-efficient infrastructure.
- Perform long-context training on standard GPUs by combining optimization techniques.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you are fine-tuning large language models and want to reduce GPU memory usage and training time.
The package is actively maintained with no known vulnerabilities and integrates with the standard Hugging Face ecosystem. The LGPL-3.0-or-later license is permissive for research and commercial use provided you distribute source code. Install friction is low. Verify that the specific models and hardware you plan to use are supported before committing to a large training run.
Install
unsloth-zoo on PyPI
Before you install
Low install friction with a pure Python wheel. Active maintenance with recent releases and a large repository (71439 stars). Depends on 27 runtime packages including torch, transformers, and trl, which are standard in the ML ecosystem.
Requires PyTorch and CUDA-capable GPU or compatible accelerator; on Windows, PyTorch must be pre-installed before installing unsloth-zoo.
License in practice
Licensed under LGPL-3.0-or-later (copyleft). Derivative works and modifications must be distributed under compatible terms; commercial use is permitted if source code and license are provided.
Quickstart
pip install unsloth-zoo
from unsloth import FastLanguageModel
import torch
model, tokenizer = FastLanguageModel.from_pretrained(
model_name="unsloth/model",
load_in_4bit=True,
)
model = FastLanguageModel.get_peft_model(
model,
target_modules=["q_proj", "v_proj"],
bias="none",
use_gradient_checkpointing="unsloth",
)Verify before relying
- Exact speedup and memory reduction percentages claimed in description for specific model/hardware combinations.
- Whether all listed model families (gpt-oss, Gemma 3n, Qwen3, Llama 4, Mistral) are fully supported in version 2026.8.12.
- Compatibility with non-NVIDIA GPUs beyond the Blackwell mention.
- Specific LoRA configuration parameters and default values supported by this version.
Package facts
| License | LGPL-3.0-or-later copyleft |
| Python support | Supports the current Python release <3.15,>=3.9 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 27 packagestorchtorchaotritontyroacceleratetrlpeftcut_cross_entropymlxmlx-lmmlx-vlmpackagingtransformersdatasetssentencepiecetqdmpsutilwheelnumpyprotobufhuggingface_hubhf_transferpillowregexmsgspectyping_extensionsfilelock |
| Maintenance | Actively maintained 0 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 1,352,315 / month, #4,012 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Programming Language :: Python |
Evidence: unsloth_zoo-2026.8.12-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “reduce vram gpu training”
- unsloth-zooUnsloth Zoo provides utilities for fine-tuning large language models…
- unslothUnsloth accelerates training and fine-tuning of large language models…
- comfy-aimdoA PyTorch VRAM allocator that dynamically offloads model weights to…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also llamafactory · unsloth · torchao · torchtune · ms-swift · llama-models · peft · nvidia-modelopt · liger-kernel · transformer-engine