============================================================
Precision Comparison: CPU vs NPU
============================================================
Max errors: sum=3.05e-05, mean=2.98e-08, std=5.96e-08
PASS: NPU precision within thresholds
============================================================
PRECISION TEST PASSED
============================================================
指标
阈值
实测值
状态
max_error_sum
< 1e-3
3.05e-05
✅ 通过
max_error_mean
< 1e-5
2.98e-08
✅ 通过
max_error_std
< 1e-5
5.96e-08
✅ 通过
输出示例
2026-05-11 08:25:46,892 - INFO - VTP-Large Vision Encoder Ascend NPU Inference
2026-05-11 08:25:51,990 - INFO - Model loaded and moved to npu:0!
2026-05-11 08:25:52,187 - INFO - Using random input tensor (256x256)...
2026-05-11 08:25:52,192 - INFO - Input shape: torch.Size([1, 3, 256, 256])
2026-05-11 08:25:52,192 - INFO - Running inference...
2026-05-11 08:25:58,426 - INFO - Features shape: torch.Size([1, 64])
2026-05-11 08:25:58,427 - INFO - Inference time: 6233.68 ms
2026-05-11 08:25:58,696 - INFO - Features (first 10): [-0.06340054 -0.03150354 ...]
2026-05-11 08:25:58,697 - INFO - Inference completed successfully!