pythonfix
bitsandbytes 0.49.2
k-bit optimizers and matrix multiplication routines.
k-bit optimizers and matrix multiplication routines.
Stars: 7954, Watchers: 7954, Forks: 822, Open Issues: 80The bitsandbytes-foundation/bitsandbytes repo was created 4 years ago and the last code push was 3 hours ago.
The project is extremely popular with a mindblowing 7954 github stars!
How to Install bitsandbytes
You can install bitsandbytes using pip
pip install bitsandbytes
or add it to a project with poetry
poetry add bitsandbytes
Package Details
- Author
- None
- License
- None
- Homepage
- None
- PyPi:
- https://pypi.org/project/bitsandbytes/
- GitHub Repo:
- https://github.com/TimDettmers/bitsandbytes
Classifiers
- Scientific/Engineering/Artificial Intelligence
Related Packages
Errors
A list of common bitsandbytes errors.
Code Examples
Here are some bitsandbytes code examples and snippets.
GitHub Issues
The bitsandbytes package has 80 open issues on GitHub
- Failed to quant MoE models with fused expert weights in transformers v5
- RuntimeError: Configured CUDA binary not found
- Ascend 910B4 qlora,Error:importlib.metadata.packagenot founderror:no package metadata was found for bitsandbytes
- Adding support for building for AMD on Windows
- Support Windows for ROCm Builds
- fix ROCm GPU architecture detection failed on windows
-
70B 4-bit LLM decode bottlenecked by HIP kernel (
kgemm_4bit_inference_naive) efficiency — 49% vs 91% memory bandwidth on ROCm/gfx1151 - Renovate pre-commit + CI a bit
- Refactor: Use logging module for diagnostics and warnings instead of print statements
- Create unistd.h
- Update cuda_specs.py
- Add mps backend (python only)
- Enable Ascend NPU Backend with Custom Ops Integration for NF4 Support
- [AMD GPU installation] The Rocm-bitsandbytes installation issues
- Support 8bit Optimizers on CPU
Related Packages & Articles
farm-haystack 1.26.4.post0
LLM framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data.
datasets 4.5.0
HuggingFace community-driven open-source library of datasets
transformers 5.2.0
Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
nlp 0.4.0
HuggingFace/NLP is an open library of NLP datasets.
beir 2.2.0
A Heterogeneous Benchmark for Information Retrieval
autogluon 1.5.0
Fast and Accurate ML in 3 Lines of Code
pennylane 0.44.0
PennyLane is a cross-platform Python library for quantum computing, quantum machine learning, and quantum chemistry. Train a quantum computer the same way as a neural network.
torchbearer 0.5.5
A model training library for pytorch
skorch 1.3.1
scikit-learn compatible neural network library for pytorch