diffusers-ascend-weight-prep

A complete toolkit for Huawei Ascend NPU model conversion and inference, supporting ONNX to .om format conversion via ATC, Python inference with ais_bench, precision comparison, and end-to-end YOLO inference.

A command reference skill for Huawei Ascend NPU npu-smi tool, covering device queries (health, temperature, power, memory), configuration, firmware upgrades, virtualization (vNPU), and certificate management.

HCCL (Huawei Collective Communication Library) performance testing skill for Ascend NPU clusters, covering distributed communication bandwidth testing, HCCL functionality verification, and collective operations benchmarking.

Streamlines Docker container creation, optimization, and deployment for AI/ML workloads on Ascend hardware with performance-focused configurations.

Optimizes machine learning model size and inference speed through advanced pruning, quantization, and compression techniques for efficient deployment.

Benchmarks and profiles AI model performance across hardware, measuring inference speed, memory usage, and computational efficiency.

Skill

Skill

Skill

Skill