bitsandbytes

bitsandbytes enables accessible large language models via k-bit quantization for PyTorch. We provide three main features for dramatically reducing memory consumption for inference and training:

• 8-bit optimizers uses block-wise quantization to maintain 32-bit performance at a small fraction of the memory cost.
• LLM.int8() or 8-bit quantization enables large language model inference with only half the required memory and without any performance degradation. This method is based on vector-wise quantization to quantize most features to 8-bits and separately treating outliers with 16-bit matrix multiplication.
• QLoRA or 4-bit quantization enables large language model training with several memory-saving techniques that don't compromise performance. This method quantizes a model to 4-bits and inserts a small set of trainable low-rank adaptation (LoRA) weights to allow training.

The library includes quantization primitives for 8-bit & 4-bit operations, through bitsandbytes.nn.Linear8bitLt and bitsandbytes.nn.Linear4bit and 8-bit optimizers through bitsandbytes.optim module.

System Requirements

bitsandbytes has the following minimum requirements for all platforms:

• Python 3.10+
• PyTorch 2.4+
  • Note: While we aim to provide wide backwards compatibility, we recommend using the latest version of PyTorch for the best experience.

Accelerator support:

Note: this table reflects the status of the current development branch. For the latest stable release, see the document in the 0.50.0 tag.

Legend:

🚧 = Planned | 〰️ = Partially Supported | ✅ = Supported | ❌ = Not Supported

Platform	Accelerator	Hardware Requirements	LLM.int8()	QLoRA 4-bit	8-bit Optimizers
🐧 Linux, glibc >= 2.24
x86-64	◻️ CPU	Minimum: AVX2 Optimized: AVX512F, AVX512BF16	✅	✅	✅
	🟩 NVIDIA GPU cuda	SM60+ minimum SM75+ recommended	✅	✅	✅
	🟥 AMD GPU cuda	CDNA: gfx908, gfx90a, gfx942, gfx950, gfx1250 RDNA: gfx101X, gfx103X, gfx110X, gfx115X, gfx120X	✅	✅	✅
	🟦 Intel GPU xpu	Data Center GPU Max Series Arc A-Series (Alchemist) Arc B-Series (Battlemage)	✅	✅	✅
	🟪 Intel Gaudi hpu	Gaudi2, Gaudi3	✅	〰️	❌
aarch64	◻️ CPU		✅ *	✅	✅
	🟩 NVIDIA GPU cuda	SM75+	✅	✅	✅
🪟 Windows 11 / Windows Server 2022+
x86-64	◻️ CPU	AVX2	✅	✅	✅
	🟩 NVIDIA GPU cuda	SM60+ minimum SM75+ recommended	✅	✅	✅
	🟥 AMD GPU cuda	CDNA: gfx908, gfx90a RDNA: gfx101X, gfx103X, gfx110X, gfx115X, gfx120X	✅	✅	✅
	🟦 Intel GPU xpu	Arc A-Series (Alchemist) Arc B-Series (Battlemage)	✅	✅	✅
arm64	◻️ CPU		✅	✅	✅
	🟩 NVIDIA GPU cuda	SM121	✅	✅	✅
🍎 macOS 14+
arm64	◻️ CPU	Apple M1+	✅ *	✅	✅
	⬜ Metal mps	Apple M1+	✅ *	✅	🚧

^{* While supported, these marked features may lack in performance optimizations.}

:book: Documentation

• Official Documentation
• 🤗 Transformers
• 🤗 Diffusers
• 🤗 PEFT

:heart: Sponsors

The continued maintenance and development of bitsandbytes is made possible thanks to the generous support of our sponsors. Their contributions help ensure that we can keep improving the project and delivering valuable updates to the community.

License

bitsandbytes is MIT licensed.

How to cite us

If you found this library useful, please consider citing our work:

QLoRA

@article{dettmers2023qlora,
  title={Qlora: Efficient finetuning of quantized llms},
  author={Dettmers, Tim and Pagnoni, Artidoro and Holtzman, Ari and Zettlemoyer, Luke},
  journal={arXiv preprint arXiv:2305.14314},
  year={2023}
}

LLM.int8()

@article{dettmers2022llmint8,
  title={LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale},
  author={Dettmers, Tim and Lewis, Mike and Belkada, Younes and Zettlemoyer, Luke},
  journal={arXiv preprint arXiv:2208.07339},
  year={2022}
}

8-bit Optimizers

@article{dettmers2022optimizers,
  title={8-bit Optimizers via Block-wise Quantization},
  author={Dettmers, Tim and Lewis, Mike and Shleifer, Sam and Zettlemoyer, Luke},
  journal={9th International Conference on Learning Representations, ICLR},
  year={2022}
}