Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
|
|
1 year ago | |
|---|---|---|
| .gitee | 1 year ago | |
| docs | 1 year ago | |
| models | 1 year ago | |
| .gitignore | 1 year ago | |
| CONTRIBUTING.md | 1 year ago | |
| LICENSE | 1 year ago | |
| README.md | 1 year ago | |
| RELEASE.md | 1 year ago | |
DeepSparkInference推理模型库作为DeepSpark开源社区的核心项目,于2024年3月正式开源,一期甄选了48个推理模型示例,涵盖计算机视觉,自然语言处理,语音识别等领域,后续将逐步拓展更多AI领域。
DeepSparkInference中的模型提供了在国产推理引擎IGIE或IxRT下运行的推理示例和指导文档,部分模型提供了基于国产通用GPU智铠100的评测结果。
IGIE(Iluvatar GPU Inference Engine)是基于TVM框架研发的高性能、高通用、全流程的AI推理引擎。支持多框架模型导入、量化、图优化、多算子库支持、多后端支持、算子自动调优等特性,为推理场景提供易部署、高吞吐量、低延迟的完整方案。
IxRT(Iluvatar CoreX RunTime)是天数智芯自研的高性能推理引擎,专注于最大限度发挥天数智芯通用GPU 的性能,实现各领域模型的高性能推理。IxRT支持动态形状推理、插件和INT8/FP16推理等特性。
DeepSparkInference将按季度进行版本更新,后续会逐步丰富模型类别并拓展大模型推理。
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| AlexNet | FP16 | Supported | Supported |
| INT8 | Supported | Supported | |
| CLIP | FP16 | Supported | - |
| INT8 | - | - | |
| Conformer-B | FP16 | Supported | - |
| INT8 | - | - | |
| CSPResNet50 | FP16 | - | Supported |
| INT8 | - | Supported | |
| DenseNet121 | FP16 | Supported | - |
| INT8 | - | - | |
| EfficientNet-B0 | FP16 | Supported | Supported |
| INT8 | - | Supported | |
| EfficientNet_B1 | FP16 | - | Supported |
| INT8 | - | Supported | |
| GoogLeNet | FP16 | Supported | Supported |
| INT8 | Supported | Supported | |
| HRNet-W18 | FP16 | Supported | - |
| INT8 | - | - | |
| InceptionV3 | FP16 | Supported | - |
| INT8 | Supported | - | |
| MobileNetV2 | FP16 | Supported | Supported |
| INT8 | Supported | Supported | |
| MobileNetV3 | FP16 | - | Supported |
| INT8 | - | - | |
| RepVGG | FP16 | - | Supported |
| INT8 | - | - | |
| Res2Net50 | FP16 | - | Supported |
| INT8 | - | - | |
| ResNet101 | FP16 | - | Supported |
| INT8 | - | - | |
| ResNet18 | FP16 | Supported | Supported |
| INT8 | Supported | Supported | |
| ResNet34 | FP16 | - | Supported |
| INT8 | - | Supported | |
| ResNet50 | FP16 | Supported | Supported |
| INT8 | Supported | - | |
| ResNeXt50_32x4d | FP16 | Supported | - |
| INT8 | Supported | - | |
| ShuffleNetV1 | FP16 | - | Supported |
| INT8 | - | - | |
| SqueezeNet 1.0 | FP16 | - | Supported |
| INT8 | - | Supported | |
| Swin Transformer | FP16 | Supported | - |
| INT8 | - | - | |
| VGG16 | FP16 | Supported | Supported |
| INT8 | Supported | - |
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| RetinaNet | FP16 | Supported | - |
| INT8 | - | - | |
| YOLOv3 | FP16 | Supported | - |
| INT8 | Supported | - | |
| YOLOv4 | FP16 | Supported | - |
| INT8 | Supported | - | |
| YOLOv5 | FP16 | Supported | - |
| INT8 | Supported | - | |
| YOLOv6 | FP16 | Supported | - |
| INT8 | - | - | |
| YOLOv7 | FP16 | Supported | - |
| INT8 | Supported | - | |
| YOLOv8 | FP16 | Supported | - |
| INT8 | Supported | - | |
| YOLOX | FP16 | Supported | Supported |
| INT8 | Supported | Supported |
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| Mask R-CNN | FP16 | - | Supported |
| INT8 | - | - |
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| FastReID | FP16 | Supported | - |
| INT8 | - | - | |
| DeepSort | FP16 | Supported | - |
| INT8 | Supported | - |
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| BERT Base NER | FP16 | - | - |
| INT8 | Supported | - | |
| BERT Base SQuAD | FP16 | Supported | Supported |
| INT8 | - | - | |
| BERT Large SQuAD | FP16 | Supported | Supported |
| INT8 | Supported | Supported |
| Models | Precision | IGIE | IxRT |
|---|---|---|---|
| Conformer | FP16 | Supported | - |
| INT8 | - | - |
请参见 DeepSpark Code of Conduct on Gitee or on GitHub。
请联系 contact@deepspark.org.cn。
请参见 DeepSparkInference Contributing Guidelines。
本项目许可证遵循Apache-2.0。
No Description
Python C++ Shell Perl SVG other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》