AKG (Auto Kernel Generator) is an optimizer for operators in Deep Learning Networks, which provides the ability to automatically fuse ops with specific patterns.
Skill Claude CodeCodex
A collection of Triton Ascend performance techniques for choosing tile sizes, loading data safely, broadcasting values, and performing reductions.
Skill Claude CodeCodex
A Triton Ascend example that runs two kernels in sequence, storing the first calculation in a temporary buffer before the second reads it.
Skill Claude CodeCodex
A complete Triton Ascend example of LayerNorm, a method that centers and scales values using their mean and variance.
Skill Claude CodeCodex
A complete example of tiled matrix multiplication in Triton for Ascend hardware. Matrix multiplication combines rows and columns of number tables and is used throughout machine learning.
Skill Claude CodeCodex
A complete example of implementing ReLU in Triton for Ascend hardware. ReLU is a common machine-learning operation that replaces negative values with zero.
Skill Claude CodeCodex
A complete example of implementing softmax in Triton for Ascend hardware. Softmax turns a group of numbers into values that add up to one, often for selecting or weighting options in machine-learning models.
Skill Claude CodeCodex
An integration example for using Triton Ascend kernels with MindSpore, a machine-learning framework. It shows how custom operators are registered and how tensors are passed into and out of them.
Skill Claude CodeCodex
Triton Ascend hard API restrictions and forbidden syntax. MUST-follow rules that apply to every kernel: forbidden control flow (return/break/continue/lambda/while), tensor slice/index restrictions, scalar conversion rules, BLOCKSIZE upper bound. Violating any of these produces a compile or runtime error on Ascend.
Skill Claude CodeCodex
A reference for Ascend hardware limits and compiler restrictions that affect Triton kernels. It explains storage areas such as UB and L0 memory, which hold data during vector and matrix operations.
Skill Claude CodeCodex
A beginner's guide to the main building blocks of Triton Ascend programming, including program IDs, blocks, grids, kernel functions, and boundary masks. Triton is a language for writing GPU- or accelerator-focused numerical kernels.
Skill Claude CodeCodex
A troubleshooting checklist for Triton kernels running on Ascend hardware. It covers compilation errors, runtime failures, incorrect results, memory access, launch settings, concurrency, and performance checks.
Skill Claude CodeCodex
A guide to choosing the grid and block settings that divide Triton work across Ascend processor cores. A grid describes the launched work, while a block is one portion of the data handled by that work.
Skill Claude CodeCodex
A guide to improving memory access in Triton kernels for Ascend hardware. It covers temporary buffer use, block sizes, aligned transfers, contiguous data, 2D block pointers, and prefetching.
Skill Claude CodeCodex
A general performance guide for Triton kernels on Ascend processors. It covers block sizes, grid layouts, core selection, memory alignment, automatic tuning, precision, and reduction patterns.
Skill Claude CodeCodex
An optimization guide for Transformer attention, which turns query, key, and value tensors into context-aware outputs. It explains blocked computation, causal masking, and online softmax for Flash Attention.
Skill Claude CodeCodex
A guide for writing Triton kernels that combine per-element calculations with a reduction, such as adding or averaging values into a smaller result. Triton is a language for writing GPU kernels.
Skill Claude CodeCodex
An optimization guide for element-wise operations, where each output value depends on the matching input value. It covers operations such as ReLU, sigmoid, arithmetic, casting, clamping, and copying on Ascend devices.
Skill Claude CodeCodex
An optimization guide for matrix multiplication, the operation that combines rows and columns of numbers to produce a new matrix. It explains tiling, matrix-compute cores, memory limits, alignment, and core selection on Ascend devices.
Skill Claude CodeCodex
An optimization guide for reduction operations, which combine many values into fewer values, such as sum, mean, maximum, softmax, or normalization.
Skill Claude CodeCodex
A troubleshooting note for a Triton CUDA math-function error. It explains that some functions, such as tanh, must be called through CUDA's libdevice interface.
Skill Claude CodeCodex
A reference for Triton CUDA's core programming functions and their parameters. It explains how to identify program blocks, load and store memory, reduce values, perform dot products, and use atomic updates.
Skill Claude CodeCodex
An implementation guide for attention, the Transformer operation that compares queries with keys and combines their values. It covers Flash Attention, which processes attention in blocks instead of storing the full sequence-by-sequence matrix.
Skill Claude CodeCodex
A beginner guide to writing Triton kernels for CUDA GPUs. It explains how GPU programs split data into blocks and use memory and threads to run work in parallel.
Skill Claude CodeCodex
A troubleshooting checklist for Triton CUDA kernels, covering compilation failures, runtime errors, incorrect results, and performance problems. It includes checks for memory bounds, indexing, shapes, concurrency, and launch settings.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: