b10-kernel
Baseten Kernel Library - High-performance GPU kernels for AI inference workloads.
Installation
From PyPI
pip install b10-kernel
Requirements:
- Python >= 3.12
- CUDA-compatible GPU and drivers
- PyTorch >= 2.8.0 with CUDA support
From Source
git clone <repository>
cd mp/kernels/b10-kernel
pip install -e .
For Development
# Install with test dependencies
pip install -e .[test]
# Install with all development dependencies
pip install -e .[dev]
Development guide
- Build the library from source
make build
make rebuild
- Run unit tests
make test
- Format code
make format
Kernel Development Workflow
Steps to add a new kernel:
- Implement the kernel in
csrc - Expose the interface in
include/b10_kernel_ops.h - Create torch extension in
csrc/common_extension.cc - Update
CMakeLists.txtto include new CUDA source - Expose Python interface in
python/b10_kernel/xxx.pyandpython/b10_kernel/__init__.py - Add unit test for the kernel in
test/test_xxx.py - Add benchmark script for the kernel in
benchmark/bench_xxx.py - Format code with
make format
PyPI Release
# update to expected version
make update_version VERSION=0.1.x
make build
make rename_wheel
make upload_wheel
Metadata
Release files for b10-kernel 0.1.6
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| b10_kernel-0.1.6-py312-none-manylinux2014_x86_64.whl | Python 3.12 | none | Linux glibc 2.17+ x86-64 | Details for b10_kernel-0.1.6-py312-none-manylinux2014_x86_64.whl |
Release files / b10_kernel-0.1.6-py312-none-manylinux2014_x86_64.whl
| Download URL | b10_kernel-0.1.6-py312-none-manylinux2014_x86_64.whl |
|---|---|
| Size | 537.7 kB |
| Tags | Linux glibc 2.17+ x86-64 Python 3.12 |
|
SHA-256 checksum How to use checksums |
e9eb5dab64cbe47e759583c69310863c5534e18684432a7741dde64369a79d98
|
|
BLAKE2b-256 checksum How to use checksums |
29a7ed37c801b7e3d636a4a608a646586fa94bab47a53511063dc08a5da505a9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Security for b10-kernel 0.1.6
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.