-
### 1. [cmrc2019](https://github.com/ExpressAI/DataLab/tree/main/datasets/cmrc2019)
* dataset = load_dataset("cmrc2019")
* task_type: cloze-multiple-choice
![image](https://user-images.githubuserco…
-
有个问题想请教大神,为何我跑了大大小小很多歌Dureader的项目,包括百度dureader的原项目,bleu和l-rougue从来没有超过个位数,为何你们的分数却如此之高。
-
https://sunyancn.github.io/2019/11/02/Dureader%E6%95%B0%E6%8D%AE%E9%9B%86/
-
```
datasets = ['hotpotqa', '2wikimqa', 'musique', 'narrativeqa', 'qasper', 'multifieldqa_en', 'gov_report', 'qmsum', 'trec', 'samsum', 'triviaqa', 'passage_count', 'passage_retrieval_en', 'multi_new…
-
dureader基线模型的网络定义中部分代码如下:
p_vec的获得依赖于start_labels的lod,在实际infer中,是不是必须有ground_truth,如果没有的话需要怎样使用该代码
wwjjy updated
5 years ago
-
hi,您好
想请教下C-MTEB中CovidRetrieval数据原论文是哪篇呢
在链接:https://github.com/FlagOpen/FlagEmbedding/tree/master/C_MTEB 里
![image](https://github.com/FlagOpen/FlagEmbedding/assets/109325850/3223a01d-bb8a-4ffc…
-
Env: Using the docker version `registry.baidubce.com/paddlepaddle/paddle:2.3.0-gpu-cuda11.2-cudnn8`
Command: Just using `bash run_dual_encoder_train.sh` to train.
Got Error:
```
'RecomputeOptimize…
-
python3 index.py zh ../data/dureader.para test_index
RocketQA model [zh_dureader_de_v2]
Download RocketQA model [zh_dureader_de_v2]
100%|███████████████████████████████████████| 693M/693M [00:18
-
请问在C-MTP(labeled)的838K数据中,T2-Ranking、mMARCO-zh、DuReader、NLI-zh这四个数据集分别占了多少呢?
-
![2019-01-19 17 12 46](https://user-images.githubusercontent.com/33299522/51424766-83a08b00-1c0d-11e9-84aa-83a8b90fd49d.png)
loss在某一部分数据变得异常大,最后导致验证集所有分数为0. 请问,是怎么产生的呢?