Closed shuaichenchang closed 1 year ago
@shuaichenchang Regarding task-dependent PyPI libraries, maybe you can refer to this strategy to figure out a good way to introduce them?
@shuaichenchang Regarding task-dependent PyPI libraries, maybe you can refer to this strategy to figure out a good way to introduce them?
I am trying to add the required libraries for text-to-sql evaluation into setup.cfg. https://github.com/neulab/ExplainaBoard/pull/555/commits/8fa204391a69563518b47f61f96e2865cffdf0f5 But it seems not to work. Do you have any suggestions on how to add it? @pfliu-nlp @tetsuok
@shuaichenchang The configurations of the failing CI jobs need to be updated to make CI green. Namely, the following places:
CI jobs (unit-tests and integration-tests) are failing because these jobs aren't configured to install required packages with the extra, text_to_sql
. In other words, these jobs need to run pip install ."[text_to_sql]"
to install required packages for text-to-sql, but they currently run pip install .
instead.
@shuaichenchang Any update on this PR? Would you check my comment? Let me know if you have any questions.
It seems I accidentally added others' commits to my PR when I merge the main into my branch. Now the PR has too many commits. Do you know how to undo them or I may need to create a PR just for the commits I submitted? @pfliu-nlp @odashi Sorry for the inconvenience.
@shuaichenchang How about:
git checkout <your preferred commit>
then git checkout -b <new tag>
@shuaichenchang How about:
- withdraw this pull request
git checkout <your preferred commit>
thengit checkout -b <new tag>
- propose a new pull request on the new tag
@odashi Sounds great! I picked my commits and created a new branch and PR.
Add text-to-sql evaluation metrics
Overview
This PR is the first of three PRs to add text-to-sql evaluation metrics.
Details
This PR adds metrics in
explainaboard/metrics
, and test functions inintegration_tests/metric_test.py
. The metrics use the code fromexplainaboard/third_party/text_to_sql_test_suit_eval
for text-to-sql task evaluations.