Replies: 12 comments
-
可以使用om/onnx高性能推理 |
Beta Was this translation helpful? Give feedback.
-
这个不支持动态输入啊?动态输入问题怎么解决? |
Beta Was this translation helpful? Give feedback.
-
像OCR识别模型,图像宽范围从几十到几千的大小,图片缩放一个大小肯定会影响整体的精度大小,不缩放统一大小,那怎么处理模型跟预处理图片让它保持整体效果? |
Beta Was this translation helpful? Give feedback.
-
OM不支持动态shape,可以固定成yaml中推荐的size,可能会损失一定精度;如果追求精度,可以使用onnx推理,onnx支持动态shape |
Beta Was this translation helpful? Give feedback.
-
onnx推理我试过,碰到每次不同的大小图片尺寸,就会生成不同的om模型缓存,而且每次都会花费很多时间,这样跑很多图片的话就会生成几百上千个om模型缓存,这是行不通的项目上使用,就想找一个更好的方法 |
Beta Was this translation helpful? Give feedback.
-
使用onnx推理为什么会产生om模型的缓存呢 |
Beta Was this translation helpful? Give feedback.
-
你说的onnx推理是指的是使用华为npu的onnxruntime-cann推理吗?如果是这个,你能提供相关使用示例代码吗? |
Beta Was this translation helpful? Give feedback.
-
不是的,基于cpu的onnxruntime |
Beta Was this translation helpful? Give feedback.
-
那没用啊,我想要使用的是华为的NPU加速 |
Beta Was this translation helpful? Give feedback.
-
我也遇到了一样的问题,这个模型缓存虽然可以关闭生成,但是性能还是很慢。。 |
Beta Was this translation helpful? Give feedback.
-
请问现在支持310P3 npu了吗? |
Beta Was this translation helpful? Give feedback.
-
请问现在支持310P3 npu了吗? |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
🔎 Search before asking
🐛 Bug (问题描述)
ocr识别模型什么时候能够支持华为升腾310P的npu使用啊?
🏃♂️ Environment (运行环境)
ocr识别模型什么时候能够支持华为升腾310P的npu使用啊?
🌰 Minimal Reproducible Example (最小可复现问题的Demo)
ocr识别模型什么时候能够支持华为升腾310P的npu使用啊?
Beta Was this translation helpful? Give feedback.
All reactions