-
### 软件环境
* fastploy版本 1.0.7最新版
* GPU版本
- paddlepaddle:2.4.2
- paddlepaddle-gpu: 11.2
- paddlenlp: 2.5.2
- cuda 11.2
- cudnn 8.1.1
**除使用`fastploy`外,所有程序均正常**
[cudnn下载官网](http…
-
训练出现问题,训练一个epoch后CUDA利用率变为0,然后电脑内存开始上升,直到被kill。有解决办法和思路吗?
最离谱的是,我发现debug不会出现类似的情况。。。
1. PaddleSeg版本:PaddleSeg release/2.7
2. PaddlePaddle版本:paddlepaddle-gpu 2.4/2.3都有同样的问题
3. 操作系统信息:Windows10…
-
### 软件环境
```Markdown
paddle-bfloat 0.1.7
paddle2onnx 1.0.6
paddlefsl 1.1.0
paddlehub 2.3.1
paddlenlp 2.6.0rc0
paddl…
-
#### 0. 网站和截图
[https://ocr.plantree.me/](https://ocr.plantree.me/)
![image](https://github.com/ruanyf/weekly/assets/20148487/aba1f86d-e92d-4ed7-9b36-82d23764f371)
#### 1. 背景
平时自己需要,离线完成OCR…
-
1. PaddleClas版本以及PaddlePaddle版本:
PaddleClas release/2.2
PaddlePaddle 2.1.1
3. 训练环境信息:
a. 具体操作系统,Kylin V10,linux master 4.19.90-17.ky10.aarch64
b. Python版本号,Python3.7.4
c. CUDA/cuDNN版本…
-
## 环境
- 【FastDeploy版本】: 按照官方编译步骤编译的昇腾版本
- 【编译命令】官网给出的命令
- 【系统平台】: Linux aarch64(Ubuntu 18.04)
- 【硬件】: 说明具体硬件型号,Ascend910
- 【编译语言】: Python3.9
- 【补充】所需要的paddle npu环境以及paddlelite环境以及FastDeploy…
-
使用paddlex_restful
新建目标检测数据集, 显示标注异常, 无法显示标注的目标,大家有没有遇到这种情况帮忙看看
paddle版本:2.2.0
paddleX版本:2.1.0
浏览器截图
![image](https://user-images.githubusercontent.com/13159218/156099240-f68a0bcf-c5b1-4842-8f58-…
-
### 请提出你的问题
RuntimeError: Downloading from https://bj.bcebos.com/paddlenlp/models/community/PaddlePaddle/ernie_vil-2.0-base-zh/model_config.json failed with code 404!
-
在用paddlex训练模型时,加入add_negative_samples方法时,背景图数量是有限制的吗?如果有数量现状,数值是多少呢?我添加500张和200张都会报错,130张时可以正常训练,而且精度反而下降0.3左右,不知道是啥原因。还望解答,非常感谢。
-
- 版本、环境信息
1)PaddleHub和PaddlePaddle版本:paddlehub==2.0.0b,paddlepaddle-gpu==2.0.0rc1
2)系统环境:WIN10,python3.79
hub serving start - c config.json 配置文件启动失败 运行这行命令 最后是直接无返回无输出 直接就为空 像没输入命令按了一下回车.
…