ascend-migration

Analyze PyTorch model architecture for Ascend NPU compatibility. Use when examining model structure, CUDA usage, distributed training patterns, and identifying migration requirements.

Migrate CUDA code to Ascend NPU code. Use when implementing PyTorch to NPU migration, replacing CUDA APIs with torch_npu equivalents.

Analyze computation-intensive operators and performance for Ascend NPU. Use when examining model operations, performance bottlenecks, and CANN operator library support.

Analyze Python package dependencies for Ascend NPU compatibility. Use when examining requirements.txt, setup.py, environment files, and checking for CUDA-dependent packages.

Analyze memory access patterns and optimization opportunities for Ascend NPU. Use when examining data loading, host-device transfers, mixed precision training, and memory efficiency.

Migrate PyTorch model training code for Ascend NPU. Use when converting training loops, optimizers, data loading, and distributed training from CUDA to NPU.