A setup workflow for starting an AscendC operator project, where an operator is a compute kernel for a Huawei Ascend AI processor. It creates a standard, compilable project skeleton that can be extended with real implementation code.
A test-case design guide for AscendC operators, which are custom computing operations for Huawei Ascend AI processors. It focuses on creating tests for accuracy, performance, general inputs, and unit-level behavior.
Complete toolkit for Huawei Ascend NPU model conversion and end-to-end inference adaptation. Workflow 1 auto-discovers input shapes and parameters from user source code. Workflow 2 exports PyTorch models to ONNX. Workflow 3 converts ONNX to .om via ATC with multi-CANN version support. Workflow 4 adapts the user's full…
An installation guide for Huawei Ascend CANN, a software toolkit for running applications on Ascend neural-processing hardware. It covers the CANN Toolkit, kernels, and NNAL components.
An installation guide for CANN, Huawei’s software platform for running artificial-intelligence workloads on Ascend NPUs. It covers setting up the Ascend environment and addressing installation problems.
A code-generation tool for Catlass operators, which are custom computing operations for Ascend NPU hardware. It turns an operator design document into the files needed for an operator project.
A guide for turning requirements for CATLASS operators into a concrete design document. CATLASS is a development framework for creating operators, which are units of computation used by software and hardware systems.
A workflow for developing Catlass operators, which are custom computing operations for Ascend NPU hardware. It connects design, code generation, compiler use, documentation, accuracy checks, and performance work.
A guide for improving the speed of Catlass operators on Ascend hardware. It uses profiler measurements, which show where a program spends time, to compare changes to tiling, the way work is divided for execution.
Ascend NPU model migration suite. Invoke when user wants to migrate/train models on Ascend NPU, setup environment, or deploy models from open-source repositories.
An installation guide for OpenMMLab computer-vision libraries on Ascend NPUs, Huawei processors designed for artificial-intelligence workloads. It covers libraries such as MMCV, MMDetection, MMDetection3D, and Detectron2, for local and remote development.
An installation procedure for mmcv-full, a computer-vision library used by OpenMMLab projects, in an Ascend NPU container. It builds the library with support for NPU operators.
An installer for MMDetection inside an Ascend NPU container, an isolated software environment. MMDetection is a computer-vision library for finding and locating objects in images.
An installation procedure for mmdetection3d, an open-source toolkit for detecting objects in three-dimensional data, inside an Ascend NPU container. It also installs its mmsegmentation dependency for image segmentation.
A toolkit for developing on remote machines over SSH, the standard secure protocol for logging into another computer. It covers connections, command execution, file transfer, deployment, network tunnels, and debugging.
A toolkit for managing SSH connections, where SSH is a way to access another computer securely over a network. It also covers remote commands, file transfers, and background tasks.
A structured troubleshooting process for problems on remote servers. It checks the environment with awareness of the current context and supports debugging inside containers, which are isolated software environments.
A tool for managing long-running coding tasks, with saved checkpoints and the ability for an agent to pause and resume. The description does not provide more implementation details.