ppc64le/linux/: fastsafetensors-0.4.0 metadata and description

High-performance safetensors model loader

author_email	Takeshi Yoshimura <[email protected]>
classifiers	Programming Language :: Python :: 3.10 Programming Language :: Python :: 3.11 Programming Language :: Python :: 3.12 Programming Language :: Python :: 3.13 Programming Language :: Python :: 3.14 Environment :: MetaData :: IBM Python Ecosystem
description_content_type	text/markdown
dynamic	license-file
keywords	fastsafetensors,safetensors,GDS
license	Apache-2.0
license_file	LICENSE
maintainer_email	Takeshi Yoshimura <[email protected]>
metadata_version	2.4
project_urls	Repository, https://github.com/foundation-model-stack/fastsafetensors
provides_extras	progress test dev threefs
requires_dist	typer>=0.9.0 tqdm>=4.66.3; extra == "progress" torch>=2.10.0; extra == "test" pytest>=9.0.3; extra == "test" pytest-cov>=5.0.0; extra == "test" transformers>=5.0.0; extra == "test" safetensors>=0.4.0; extra == "test" pre-commit>=3.6.0; extra == "dev" black==26.3.1; extra == "dev" isort==6.1.0; extra == "dev" flake8==7.3.0; extra == "dev" mypy==1.19.1; extra == "dev" fastsafetensor-3fs-reader>=0.3.3; extra == "threefs"
requires_python	<3.15,>=3.10

File	Tox results	History
fastsafetensors-0.4.0-cp312-cp312-manylinux_2_39_ppc64le.whl Size 275 KB Type Python Wheel Python 3.12		Uploaded to ppc64le/linux by ppc64le 2026-09-29 07:06:56
fastsafetensors-0.4.0-cp313-cp313-manylinux_2_39_ppc64le.whl Size 2 MB Type Python Wheel Python 3.13		Uploaded to ppc64le/linux by ppc64le 2026-09-29 07:06:57
fastsafetensors-0.4.0-cp314-cp314-manylinux_2_39_ppc64le.whl Size 276 KB Type Python Wheel Python 3.14		Uploaded to ppc64le/linux by ppc64le 2026-09-29 07:06:58

fastsafetensors

fastsafetensors is an efficient safetensors loader. If you develop your own code that loads large safetensors files, you can try fastsafetensors APIs (see docs). For example, vLLM and SGLang have --load-format fastsafetensors command-line argument to speed up their initialization.

This library supports Linux/CUDA, ROCm, Windows, 3FS, unified-memory systems such as DGX Spark, and so on. We welcome more platform/storage-specific optimizations like them by adding new copier backends. Our CI tests Python 3.10-3.14 with PyTorch 2.11.0.

Performance Highlights

Performance highlights from the CLOUD 2025 paper and benchmark docs:

Standalone model loading was 4.8x-7.5x faster than the default safetensors deserializer on Llama, Falcon, and Bloom models, and reached 26.4 GB/s NVMe read throughput for Llama-70B on four GPUs with GDS.
In the paper's vLLM integration experiment, startup time dropped from 12.39s to 4.74s for Llama-2-13B on 4x L40S GPUs, and from 16.04s to 6.88s on 1x A100.
On AMD ROCm without GDS, the documented nogds path reached 6.02 GB/s for GPT-2 Medium versus 1.28 GB/s with mmap (4.7x throughput), and 2.62 GB/s for GPT-2 versus 1.01 GB/s with mmap (2.6x throughput). See the report for more details.

Quick Start

pip install fastsafetensors
pip install vllm # for quick demo
vllm serve Qwen/Qwen3-0.6B --load-format fastsafetensors
...
Loading safetensors using Fastsafetensor loader:   0% Completed | 0/1 [00:00<?, ?it/s]
Loading safetensors using Fastsafetensor loader: 100% Completed | 1/1 [00:00<00:00,  1.23it/s]

Design Details

See Overview for features, basic API usage, and configuration.

Code of Conduct

Please refer to Foundation Model Stack Community Code of Conduct.

Development

See Development.

Publication

Takeshi Yoshimura, Tatsuhiro Chiba, Manish Sethi, Daniel Waddington, Swaminathan Sundararaman. (2025) Speeding up Model Loading with fastsafetensors arXiv:2505.23072 and IEEE CLOUD 2025.

devpi