xujiashuai/timm-tf_efficientnetv2_b1.in1k-FE-NPU
Model Introduction
Files & versions
Pull Requests
Discussion
Analysis

timm/tf_efficientnetv2_b1.in1k on Ascend NPU

1. 简介

  • 模型来源: timm/tf_efficientnetv2_b1.in1k
  • 适配状态: SUCCESS
  • 适配时间: 2026-05-20

2. 验证环境

组件版本
torch2.9.0
torch-npu2.9.0.post1
CANN8.5.1
NPUAscend 910B4
OSUbuntu 22.04.5 LTS (aarch64)
Python3.11.14

3. 精度评测

NPU vs CPU 精度对比

指标数值
Cosine Similarity1.000000
Max Absolute Error0.003107
是否满足要求是(< 1%)

4. 性能

指标数值
平均延迟9.86 ms
测试轮数10

5. 推理输出证据

=== tf_efficientnetv2_b1.in1k ===
Params: 8,141,052

--- CPU ---
Shape: [1, 1000]
Prediction: 539

--- NPU ---
Shape: [1, 1000]
Prediction: 539

--- Comparison ---
Cosine Similarity: 1.000000
Max Abs Error: 0.003107
CPU pred: 539, NPU pred: 539

--- Latency ---
Avg: 9.86 ms (10 runs)

Done.

6. 复现步骤

# 环境准备
source /usr/local/Ascend/ascend-toolkit/set_env.sh

# 运行推理
/usr/local/python3.11.14/bin/python3 inference.py --device npu:0

7. 评测材料

材料文件
推理脚本inference.py
精度报告report.json
截图证据screenshots/