Together
Official@togethercomputer
Offers high-performance kernel optimization and model integration services for SGLang-based diffusion and generative architectures.
Agent Skills by Together
Showing 6 vetted skills indexed across 1 GitHub repositories.
togethercomputeradd-new-diffusion-model
Guide end-to-end integration of new diffusion models into SGLang pipelines.
togethercomputerdiffusion-optimal-perf
Tune CLI flags and offload strategies to reduce latency and VRAM usage in SGLang-Diffusion.
togethercomputerdiffusion-kernel
Build, test, and profile JIT CUDA and Triton kernels for SGLang diffusion models.
togethercomputerwrite-sglang-test
Automate SGLang CI/UT test setup with standardized templates and fixtures.
togethercomputeradd-jit-kernel
Implement a lightweight JIT CUDA kernel for SGLang's jit_kernel module.
togethercomputeradd-sgl-kernel
Integrate an AOT CUDA/C++ kernel with Torch bindings and CMake.
Frequently Asked Questions About Together
FAQPage SchemaWhat specific tasks can be performed using Together's kernel expertise?▼
Together enables the integration of new diffusion models into SGLang, the development of JIT CUDA and Triton kernels, and the implementation of AOT C++ kernels with Torch bindings. These capabilities focus on reducing latency and optimizing VRAM usage for high-performance generative model serving.
Which technical personas benefit from these SGLang-focused capabilities?▼
These skills are designed for machine learning engineers, kernel developers, and infrastructure architects working on high-throughput generative model deployment. Professionals focused on low-level performance tuning and custom model integration within the SGLang ecosystem will find these technical resources essential for production-grade inference.
What are the prerequisites for implementing these kernel and model integrations?▼
Implementation requires an existing SGLang environment, proficiency in CUDA and Triton programming, and familiarity with CMake and Torch bindings. Users must have a baseline understanding of diffusion model architectures and the specific requirements for JIT and AOT kernel compilation within the SGLang framework.