要不要支持下gpu解析 #142
Replies: 11 comments
-
|
正在写，但是onnx的gpu支持很差其实，启动的时候资源占用会比较多 |
Beta Was this translation helpful? Give feedback.
-
可以转tensorrt吗？我的场景涉及到大量并发，CPU容易成为瓶颈，所以考虑用GPU，我正在优化代码。 |
Beta Was this translation helpful? Give feedback.
-
欢迎提供pr贡献哈，理论上onnx转tensorrt也需要看算子的兼容性 |
Beta Was this translation helpful? Give feedback.
-
|
大佬现在当前表格识别是不是不支持GPU呀？ |
Beta Was this translation helpful? Give feedback.
-
表格识别的模型有没有pth呀？onnxruntime在创建大量分布在所有可用 GPU 设备上的 ONNX Runtime 会话时候，会有BUG，设置了device id实际全在一个device id上，导致OOM。 |
Beta Was this translation helpful? Give feedback.
-
|
表格模型都是onnx哈，另外一个库rapidTable有pytorch的模型 |
Beta Was this translation helpful? Give feedback.
-
|
麻烦问下现在可以支持GPU推理了吗？感谢 |
Beta Was this translation helpful? Give feedback.
-
gpu的话，应该可以用，但是onnx的gpu推理其实也没有明显提速的，可以试试哈，需要安装onnxruntime-gpu吧，参数上也要指定使用gpu |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
我用起来确实加速了，只不过收益不大，后续请关注我的RapidOCR-pytorch项目，使用pth模型和pytorch支持，接口和RapidOCR几乎一样，且误差和paddle版本，onnx版本差别很小 |
Beta Was this translation helpful? Give feedback.
-
|
这个项目的原始模型（阿里的两个）就是PyTorch的，只需要做一下适配就可以用PyTorch来推理了。自然也就支持GPU了。 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
感觉cpu一下就炸了
Beta Was this translation helpful? Give feedback.
All reactions