-
### 请提出你的问题 Please ask your question
看到官方说可以将retriever中的默认embedding模型换成自己的模型,我先尝试了TaskFlow里自带的embedding模型比如rocketqa-zh-dureader-cross-encoder等,是没有问题的。
但是,当我自己下载了其他模型,模型路径里面也包含了基本的模型文件后,将query_embedd…
-
paddle-bfloat 0.1.7
paddle-pipelines 0.5.0
paddle2onnx 0.9.6
paddlefsl 1.1.0
paddlenlp …
-
请问提到的multi-task是怎么样的一种思路,想学习参考一下
-
I saw
https://github.com/PaddlePaddle/RocketQA/blob/main/research/DuReader-Retrieval-Baseline/metric/evaluation.py
[](https://github.com/PaddlePaddle/RocketQA/blob/main/research/DuReader-Retrieval-…
-
试了一些数据增强方法,自己构造不可回答样例,然而并没啥用,可能打开方式还不对
-
训练文件 example.py
报错信息
Traceback (most recent call last):
File "/opt/qa/RocketQA/examples/example.py", line 66, in
train_cross_encoder('zh_dureader_ce_v2', './data/cross.train.tsv')
File "…
-
不好意思。 打扰了。 后面的bert训练知道 但是不知道你最初为什么要来一个抽取,这里主要抽取具体是想干什么? 感觉处理过的数据貌似跟原始数据没啥差别。。
-
-
### 请提出你的问题
将语义检索系统部署好之后,使用web页面的左下角的文件(txt文件)上传功能时出错(即便使用github给出的示例数据同样报错),错误信息如下:
```shell
JSONDecodeError: Expecting value: line 1 column 1 (char 0)
Traceback:
File "/usr/local/lib/python…
-
![2019-01-17 18 13 06](https://user-images.githubusercontent.com/33299522/51311325-a141e900-1a83-11e9-9488-23c427f7ddf7.png)
Could you explain why you add self.vocab_size between question id and answ…