sgl-deep-gemm
SGLang fork of DeepGemm
Decision gist · record as of 2026-08-14
Yes, if you are using SGLang for LLM inference on NVIDIA GPUs and need optimized matrix multiplication. The active maintenance, permissive license, and ABI-compatible wheels make it low-friction to adopt. No, if you are not in the SGLang ecosystem or lack a compatible GPU and Linux platform (x86_64 or aarch64).AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python >=3.10 and NVIDIA CUDA-capable GPU; wheels available only for manylinux2014 x86_64 and aarch64 architectures.
- Medium install friction due to platform-specific wheels (manylinux2014 x86_64 and aarch64 only) and a compiled dependency on apache-tvm-ffi.
- Active maintenance with a recent release (9 days old) and steady repository activity.
License · maintenance · safety
Apache-2.0 (permissive) — Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most deployment scenarios.
last release 2026-08-05 (9 days) · last repo commit 2026-08-05 · 36 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 692,839 downloads/mo, #5,319 on PyPI
Alternatives
Verify before relying
pip install sgl-deep-gemm
import sgl_deep_gemm
# Use optimized GEMM kernels via apache-tvm-ffi bindings- What specific GEMM operations and tensor shapes the kernels optimize for.
- Performance benchmarks compared to standard CUDA GEMM implementations.
- Whether apache-tvm-ffi is automatically installed or requires separate setup.
- Compatibility with specific CUDA and cuDNN versions.
What it is and what it does
sgl-deep-gemm is an SGLang-optimized fork of DeepGemm that provides GPU-accelerated matrix multiplication kernels for neural network inference and training workloads. It wraps custom GEMM implementations via apache-tvm-ffi to enable ABI-compatible wheels that work across different Python versions without recompilation, simplifying deployment compared to building from source.
The package is designed for use within SGLang's inference framework and targets NVIDIA GPUs on Linux (x86_64 and aarch64). Installation is straightforward via pip, but it requires Python 3.10 or later and a CUDA-capable GPU. The underlying kernels are optimized for specific tensor operations common in large language model inference.
Use it for
- Accelerate matrix multiplication operations in SGLang-based LLM inference pipelines.
- Deploy optimized GEMM kernels across multiple Python versions without rebuilding wheels.
- Improve throughput for batch matrix operations on NVIDIA GPUs in production systems.
- Integrate custom GPU kernels into neural network frameworks via TVM's FFI layer.
- Support both x86_64 and aarch64 GPU deployments with a single wheel distribution.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you are using SGLang for LLM inference on NVIDIA GPUs and need optimized matrix multiplication.
The active maintenance, permissive license, and ABI-compatible wheels make it low-friction to adopt. No, if you are not in the SGLang ecosystem or lack a compatible GPU and Linux platform (x86_64 or aarch64).
Install
sgl-deep-gemm on PyPI
Before you install
Medium install friction due to platform-specific wheels (manylinux2014 x86_64 and aarch64 only) and a compiled dependency on apache-tvm-ffi. Active maintenance with a recent release (9 days old) and steady repository activity.
Requires Python >=3.10 and NVIDIA CUDA-capable GPU; wheels available only for manylinux2014 x86_64 and aarch64 architectures.
License in practice
Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most deployment scenarios.
Quickstart
pip install sgl-deep-gemm
import sgl_deep_gemm
# Use optimized GEMM kernels via apache-tvm-ffi bindingsVerify before relying
- What specific GEMM operations and tensor shapes the kernels optimize for.
- Performance benchmarks compared to standard CUDA GEMM implementations.
- Whether apache-tvm-ffi is automatically installed or requires separate setup.
- Compatibility with specific CUDA and cuDNN versions.
Package facts
| License | Apache-2.0 permissive |
| Python support | Supports the current Python release >=3.10 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 1 packageapache-tvm-ffi |
| Maintenance | Actively maintained 9 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 692,839 / month, #5,319 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Environment :: GPU :: NVIDIA CUDAProgramming Language :: Python :: 3 |
Evidence: sgl_deep_gemm-0.1.5.post2-py3-none-manylinux2014_aarch64.whl; sgl_deep_gemm-0.1.5.post2-py3-none-manylinux2014_x86_64.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 › “gpu matrix multiplication kernels”
- sgl-deep-gemmsgl-deep-gemm provides optimized GEMM (matrix multiplication) kernels…
- nvidia-cudnn-frontendProvides Python and C++ APIs to NVIDIA's cuDNN library, exposing…
- humming-kernelsHumming is a JIT-compiled GEMM kernel library for quantized matrix…
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 gram-newton-schulz · sgl-kernel · sglang-kernel · tilelang · flashinfer-python · humming-kernels · cpm-kernels · apache-tvm-ffi · nvidia-cublas · nvidia-cudnn-frontend