代码拉取完成，页面将自动刷新
SVT Base is a mid-sized variant of the Sparse Vision Transformer (SVT) series, designed to combine the expressive power of Vision Transformers (ViTs) with the efficiency of sparse attention mechanisms. By employing sparse attention and multi-stage feature extraction, SVT-Base reduces computational complexity while retaining global modeling capabilities.
| Iluvatar GPU | IXUCA SDK |
|---|---|
| MR-V100 | 4.2.0 |
Pretrained model: https://download.openmmlab.com/mmclassification/v0/twins/twins-svt-base_3rdparty_8xb128_in1k_20220126-e31cc8e9.pth
Dataset: https://www.image-net.org/download.php to download the validation dataset.
# Install libGL
## CentOS
yum install -y mesa-libGL
## Ubuntu
apt install -y libgl1-mesa-glx
pip3 install -r requirements.txt
# git clone mmpretrain
git clone -b v0.24.0 https://github.com/open-mmlab/mmpretrain.git
# export onnx model
python3 export.py --cfg mmpretrain/configs/twins/twins-svt-base_8xb128_in1k.py --weight twins-svt-base_3rdparty_8xb128_in1k_20220126-e31cc8e9.pth --output svt_base.onnx
# Use onnxsim optimize onnx model
onnxsim svt_base.onnx svt_base_opt.onnx
export DATASETS_DIR=/Path/to/imagenet_val/
# Accuracy
bash scripts/infer_svt_base_fp16_accuracy.sh
# Performance
bash scripts/infer_svt_base_fp16_performance.sh
| Model | BatchSize | Precision | FPS | Top-1(%) | Top-5(%) |
|---|---|---|---|---|---|
| SVT Base | 32 | FP16 | 673.165 | 82.865 | 96.213 |
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。
马建仓 AI 助手
