psunlpgroup / xsemplr Goto Github PK

Description:

Hello!
First of all, thank you for open-sourcing your amazing work.

I encountered some issues while trying to reproduce the results of mT5 on the mSpider dataset. (setting: monolingual training on english, which will be used for zero-shot cross-lingual evaluation on other languages). I followed your provided script located here.

Missing dependencies
In order to install the required dependencies, I followed the instructions provided here. However, I couldn't locate the "py3.7pytorch1.8.yaml" file in your repository. It seems to be missing. Could you please upload the dependency file so that I can proceed with the installation?

Missing files

1. The file mspider.py in line 93 to 113 is responsible for constructing the mSpider dataset. However, the required file for the "db_path" is missing in the repository. As a workaround, I used the original spider db provided from here. Can I use their released db? Additionally, it would be very helpful if you could provide instructions on how to handle this.

2. The files mspider.py and spider.py require importing "from third_party.miscs.bridge_content_encoder import get_database_matches," but the "miscs" folder is missing from the repository. It would be greatly appreciated if you could upload the missing folder. While searching for similar code, I found a file here in the PICARD repository. Can I use that file as a substitute? Also, It would be amazing if you could upload that file .

3. In order to evaluate the results, the configuration file used for training requires the spider evaluator, located here. However, the spider evaluator is missing from the "model/UniPSP/metric" directory. Could you please upload the spider evaluator? As a temporary solution, I tried using similar code from here, but the results are not reproducible.
   Could you upload the spider evaluator?

Thank you for your attention to these issues.