Closed suiyueliushang closed 2 years ago
We don't release ground truth for this dataset. If you're interested in, please send your submissions to us and we will give you the results.
I want to apply codeBERT to Chinese and train a Chinese code search model based on CoSQA. However I find there are not enough test data. Thank you very much if you could send the result.
We don't release ground truth for this dataset. If you're interested in, please send your submissions to us and we will give you the results. Do you mean I seed my predictions to you and you return the test results to me?
Do you mean I seed my predictions to you and you return the test results to me?
Yes. And if you need the ground truth of test set for research use, please also send your request by emails. We cannot share in GitHub.
Could you send a separated email to codexglue@microsoft.com, so that we could send files as attachment to your own email address, not in the GitHub issue (we can't know your email address here..)?
There are 1046 pieces of data in the test_webquery. But why are there only five answers in the webquery_answers? How can I get the left answers? webquery-test-1 1 webquery-test-2 1 webquery-test-3 1 webquery-test-4 0 webquery-test-5 0