-
### 请提出你的问题
### 请提出你的问题
蒸馏UIE-mini 模型底座ernie-3.0-mini-zh
:UIE蒸馏模型的部署可以直接用UIE deploy 的infer.py,? Taskflow 加了uie-data-distill-gp,只能用以下方式推理?
ie = Taskflow("information_extraction",model="uie-data-di…
-
### 请提出你的问题
蒸馏UIE-mini 模型底座ernie-3.0-mini-zh
:UIE蒸馏模型的部署可以直接用UIE deploy 的infer.py,? Taskflow 加了uie-data-distill-gp,只能用以下方式推理?
ie = Taskflow("information_extraction",model="uie-data-distill-gp", ta…
-
### 软件环境
```Markdown
- paddlepaddle: 2.5.0
- paddlepaddle-gpu: 2.4.2.post117
- paddlenlp: 2.6.0rc0
```
### 重复问题
- [X] I have searched the existing issues
### 错误描述
```Markdown
按照readme
运行du-ui…
-
请问有在公开数据集上比较过pytorch-uie和paddle-uie的结果吗?有没有具体的数据呢?
-
### 请提出你的问题
目前的UIE-X模型太大了,启动占用显存非常大。能否参考UIE的模型,给出:medium,tiny的版本的呢。急迫!
-
楼主,请问这套UIE支持嵌套实体抽取吗?
我尝试了下uie_predictor,发现无法抽出嵌套实体?
-
### 请提出你的问题
就是有没有UIE在标准数据集上面的测评。想看看UIE的效果比bert强还是弱。(以命名实体领域为例)。因为在使用UIE过程中发现,UIE抽取比bert差好多。所以想看看标准数据集上,是强还是弱。
-
![image](https://github.com/cmnfriend/O-LoRA/assets/60780329/566db127-090b-4e13-9f57-5ac205d8b8b7)
FileNotFoundError: [Errno 2] No such file or directory: 'G:/code/O-LoRA-main/O-LoRA-main/logs_and_…
-
### 问题描述 Issue Description
在910显卡上参照[官方文档2.6分支](https://github.com/PaddlePaddle/PaddleCustomDevice/blob/release/2.6/backends/npu/README_cn.md)源码编译安装了paddle框架,测试UTC模型时,发现使用npu可以正常推理,但速度很慢,30个字的单条预测居然要…
-
我有几点想反馈一下。
1.为了后期各种文本的训练,也方便人员标注。能不能以doccano或者labelstudio标准格式的数据集训练
2.能不能支持像UIE或者rexuni这种的可以自定义各种schema,然后去抽取格式。(最好兼容上面标注格式)
3.自己分段,超过512,也截断输出。把最后结果合并输出。
4.不管是实体和关系抽取,都想输出start和end,还有概率