sisrfeng
sisrfeng
### PR types Others ### PR changes Docs ### Description Fix the typo of `uvicorn`, helping to close https://github.com/PaddlePaddle/PaddleNLP/issues/7951
### 软件环境 ```Markdown - paddlepaddle: - paddlepaddle-gpu: - paddlenlp: . ``` ### 重复问题 - [X] I have searched the existing issues ### 错误描述 ```Markdown https://huggingface.co/spaces/PaddlePaddle/ERNIE-Layout 上传图片后点run, 一直显示error. ``` ### 稳定复现步骤...
### 软件环境 ```Markdown - paddlepaddle: - paddlepaddle-gpu: - paddlenlp: ``` ### 详细描述 ```Markdown https://github.com/PaddlePaddle/PaddleNLP/blob/8d6e81339a70c20b0bc96a690c4184c178c0c97d/docs/server.md?plain=1#L3 "是基于 unicorn 封装的模型部署服务化工具", 其实是nvicorn才对吧? 另外, 这份文档没提供英文版是吗? 哪些文档会给英文版呢 ```
Reason: 
thx!
The most active fork I found is: https://github.com/Uchiha-Senju/PrettyErrors_modified
``` Instead of actual stochastic oversampling, we create multiple copies of images with small objects offline for efficiency. ``` Thx!
> Quantization is currently only supported for CPUs. Someone on zhihu says: Q: pytorch量化后的模型 能使用GPU进行预测吗? A: 可以的,里面都是伪量化操作。 得到的预测结果应该还是对的,只是精度低了一点点, 会更慢 部署到FPGA/ASIC等特定硬件上才能看出加速效果