EfficientNet B6 is a high-performance convolutional neural network model developed by Google, utilizing Compound Scaling to balance the scaling of depth, width, and resolution. It incorporates Inverted Residual Blocks, Squeeze-and-Excitation (SE) modules, and the Swish activation function. EfficientNet-B6 excels in tasks like image classification and object detection. While it requires significant computational resources, its precision and efficiency make it an ideal choice for complex vision tasks.
| GPU | IXUCA SDK | Release | Branch |
|---|---|---|---|
| MR-V100 | 4.4.0 | 26.03 | release/26.03 |
| MR-V100 | 4.3.0 | 25.12 | release/25.12 |
Note: 请切换到与您的 SDK 版本对应的 Release 分支进行测试。请勿直接在 master 分支上运行测试，因为 master 分支可能包含与您的本地 SDK 版本不兼容的最新更改。
切换分支命令示例：
git checkout release/26.03
Pretrained model: https://download.pytorch.org/models/efficientnet_b6_lukemelas-24a108a5.pth
Dataset: https://www.image-net.org/download.php to download the validation dataset.
pip3 install -r ../../igie_common/requirements.txtpython3 ../../igie_common/export.py --model-name efficientnet_b6 --weight efficientnet_b6_lukemelas-24a108a5.pth --output efficientnet_b6.onnxexport DATASETS_DIR=/Path/to/imagenet_val/
export RUN_DIR=../../igie_common/# Accuracy
bash scripts/infer_efficientnet_b6_fp16_accuracy.sh
# Performance
bash scripts/infer_efficientnet_b6_fp16_performance.sh| Model | BatchSize | Precision | FPS | Top-1(%) | Top-5(%) |
|---|---|---|---|---|---|
| Efficientnet_b6 | 32 | FP16 | 523.225 | 74.388 | 91.835 |