find-knowledge / IBR

2 stars 0 forks source link

Error running the trained IBR Models #1

Open SachinVashisth opened 1 year ago

SachinVashisth commented 1 year ago

Hi

I have been trying to run the trained model natlang_model.bin on the ParaRules dataset but the error is coming as it is not able to find the file turk-questions-test-mappings.tsv in the dataset downloaded from the link provided in the README.md.

Also, in line 415 of main_natlang.py i.e.
parser.add_argument("--evaluate_model_name_or_path", default='./output/natlang/pytorch_model.bin', type=str, required=False)
isn't it should be natlang_model.bin instead of pytorch_model.bin?

find-knowledge commented 1 year ago

Hi

I apologize for not replying to the email in time, it was great to receive your email.  I have updated the data and code on github and also found an issue with the model code on the ParaRules dataset. 1) files such as turk-questions-test-mappings.tsv have been uploaded. 2) the original model code for the ParaRules dataset has also been uploaded, apologies for the oversight in our work. 3) it should be natlang_ model.bin instead of pytorch_model.bin.

If there are any more questions, you are welcome to email again. Best wishes!

 

------------------ 原始邮件 ------------------ 发件人: "find-knowledge/IBR" @.>; 发送时间: 2023年3月27日(星期一) 晚上6:06 @.>; @.***>; 主题: [find-knowledge/IBR] Error running the trained IBR Models (Issue #1)

Hi

I have been trying to run the trained model natlang_model.bin on the ParaRules dataset but the error is coming as it is not able to find the file turk-questions-test-mappings.tsv in the dataset downloaded from the link provided in the README.md.

Also, in line 415 of main_natlang.py i.e.

parser.add_argument("--evaluate_model_name_or_path", default='./output/natlang/pytorch_model.bin', type=str, required=False)

isn't it should be natlang_model.bin instead of pytorch_model.bin?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>