ascend-ai-coding
Official@ascend-ai-coding
Optimizing high-performance computing and model deployment on Huawei Ascend hardware through specialized kernel development, quantization, and distributed benchmarking.
Agent Skills by ascend-ai-coding
Showing 8 vetted skills indexed across 1 GitHub repositories.
ascend-ai-codingnpu-smi
Query and manage Huawei Ascend NPU health, configuration, and firmware via npu-smi.
ascend-ai-codingascendc
Implement AscendC transformer operator definitions and kernels for FFN, GMM, and MoE.
ascend-ai-codingascend-docker
Automate Docker container creation for Huawei Ascend NPU development with device mappings.
ascend-ai-codingmsmodelslim
Quantize and optimize Huawei Ascend NPU models for MindIE or vLLM-Ascend deployment.
ascend-ai-codingatc-model-converter
Convert ONNX models to Ascend OM format with ATC and validate via AIS Bench.
ascend-ai-codingvllm-ascend
Serve OpenAI-compatible LLM inference on Huawei Ascend NPUs.
ascend-ai-codingais-bench
Evaluate AI model accuracy and performance on Ascend NPU with AISBench.
ascend-ai-codinghccl-test
Benchmark HCCL operations like AllReduce and AllGather across Ascend NPUs with MPI orchestration.
Frequently Asked Questions About ascend-ai-coding
FAQPage SchemaWhat specific tasks can be performed using these Ascend-focused capabilities?▼
These capabilities enable the management of NPU health, conversion of ONNX models to OM format, quantization for deployment, and the execution of performance benchmarks for distributed communication operations like AllReduce and AllGather.
Which technical personas are the primary users of these Ascend-specific resources?▼
These resources are designed for hardware acceleration engineers, performance optimization specialists, and infrastructure developers tasked with deploying large-scale models on Huawei Ascend hardware environments.
What are the primary prerequisites for running these Ascend-specific operations?▼
Successful execution requires a configured Huawei Ascend NPU environment, appropriate device drivers, and the installation of the Ascend-specific runtime stack to support kernel compilation, model conversion, and distributed communication benchmarking.