FLA
Official@fla-org
Offers specialized kernel optimization and performance benchmarking for Flash Linear Attention architectures across heterogeneous GPU backends.
Agent Skills by FLA
Showing 9 vetted skills indexed across 1 GitHub repositories.
fla-orgfla-design-coverage
Defines contract-first design and coverage requirements for FLA kernel and numerical changes.
fla-orgfla-ascend-performance
Profile, diagnose, and optimize Triton-Ascend NPU kernels in the flash-linear-attention repository.
fla-orgfla-dispatch-backends
Manage and optimize backend dispatch for Flash Linear Attention operations.
fla-orgfla-optimization-loop
Optimize Flash Linear Attention kernels across Triton, Gluon, TileLang, and CuTe backends.
fla-orgfla-mr-readiness
Check pull requests against CONTRIBUTING guidelines and PR body format.
fla-orgfla-kda
Manage KDA development tasks within the Flash Linear Attention framework.
fla-orgfla-nvidia-performance
Profile and benchmark NVIDIA GPU kernels in the FLA repository.
fla-orgfla-correctness-coverage
Automate kernel correctness testing and coverage analysis in FLA/ops/.
fla-orgfla-triton-to-gluon
Port Triton kernels to Gluon with explicit tensor layout and memory control.
Frequently Asked Questions About FLA
FAQPage SchemaWhat specific tasks does FLA enable for kernel developers?▼
FLA enables developers to optimize Flash Linear Attention kernels, port Triton implementations to Gluon, and perform rigorous benchmarking on NVIDIA hardware. It provides structured mechanisms for managing KDA development tasks and ensuring kernel correctness through systematic coverage analysis and pull request validation.
Which technical personas benefit from using these capabilities?▼
These capabilities are designed for GPU kernel engineers, machine learning infrastructure researchers, and performance engineers working on linear attention architectures. It specifically targets developers focused on low-level hardware acceleration, tensor memory management, and maintaining high-performance standards within specialized neural network frameworks.
What are the primary dependencies for running FLA optimization routines?▼
Running these routines requires an environment configured for NVIDIA GPU development, including support for Triton, Gluon, TileLang, and CuTe backends. Users must ensure their local environment supports the specific tensor layout and memory control requirements defined within the FLA repository structure.