Closed tomsherborne closed 3 years ago
Hi, Do you have any plans to update the code yourself in the future? I've been trying to replicate the results from the paper but I wanted to use a more up to date Python and PyTorch version. Therefore, I've spent some time modifying the model to run with PY3.7 and Pytorch 1.2 (not >1.3.1 as I'm testing using a K40 GPU currently). If you didn't have a plan to make these changes, I have a version with minimal changes from the original code that I can submit as a PR if you'd like to use.
Hi, Thank you for your work! We would like to use your code if it can achieve the same or even better results than our original code. To this end, we are wondering if you use the official evaluation shown in our CSpider Challenge to test your code since it is easy to operate. If your code does work, we will update the scores in our challenge and show your contribution. If you have any problem with the evaluation process on codalab, feel free to contact me and i will find my co-author Yuefeng @yuefeng-shi to help you.
Thanks, it might take me some time to get through the codalab process but i'll submit a PR when it's done.
Thanks, it might take me some time to get through the codalab process but i'll submit a PR when it's done.
And you can also submit a PR now if it is ready so that we can review the code first. If you just want to submit a PR, we could test the code also. Please feel free to let us know. If you have any idea about our project, welcome to communicate with us.
Development Set Evaluation using the evaluation script in the repo. I'll proceed with the CodaLab evaluation soon.
easy medium hard extra all
count 250 440 174 170 1034
====================== EXACT MATCHING ACCURACY =====================
exact match 0.316 0.134 0.184 0.018 0.167
---------------------PARTIAL MATCHING ACCURACY----------------------
select 0.664 0.394 0.618 0.425 0.502
select(no AGG) 0.676 0.410 0.618 0.425 0.512
where 0.376 0.247 0.226 0.078 0.245
where(no OP) 0.385 0.286 0.301 0.117 0.284
group(no Having) 0.333 0.427 0.512 0.376 0.415
group 0.333 0.420 0.488 0.365 0.404
order 0.429 0.297 0.706 0.747 0.556
and/or 1.000 0.916 0.908 0.895 0.932
IUEN 0.000 0.000 0.057 0.125 0.074
keywords 0.835 0.808 0.703 0.618 0.757
---------------------- PARTIAL MATCHING RECALL ----------------------
select 0.664 0.393 0.615 0.418 0.500
select(no AGG) 0.676 0.409 0.615 0.418 0.510
where 0.380 0.253 0.228 0.061 0.237
where(no OP) 0.389 0.292 0.304 0.092 0.275
group(no Having) 0.500 0.427 0.538 0.405 0.442
group 0.500 0.420 0.513 0.392 0.431
order 0.545 0.293 0.610 0.728 0.544
and/or 0.992 0.953 0.969 0.942 0.964
IUEN 0.000 0.000 0.048 0.056 0.051
keywords 0.913 0.804 0.667 0.600 0.755
---------------------- PARTIAL MATCHING F1 --------------------------
select 0.664 0.394 0.617 0.421 0.501
select(no AGG) 0.676 0.410 0.617 0.421 0.511
where 0.378 0.250 0.227 0.069 0.241
where(no OP) 0.387 0.289 0.303 0.103 0.280
group(no Having) 0.400 0.427 0.525 0.390 0.428
group 0.400 0.420 0.500 0.378 0.417
order 0.480 0.295 0.655 0.737 0.550
and/or 0.996 0.934 0.937 0.918 0.947
IUEN 1.000 1.000 0.052 0.077 0.061
keywords 0.873 0.806 0.684 0.609 0.756
Development Set Evaluation using the evaluation script in the repo. I'll proceed with the CodaLab evaluation soon.
easy medium hard extra all count 250 440 174 170 1034 ====================== EXACT MATCHING ACCURACY ===================== exact match 0.316 0.134 0.184 0.018 0.167 ---------------------PARTIAL MATCHING ACCURACY---------------------- select 0.664 0.394 0.618 0.425 0.502 select(no AGG) 0.676 0.410 0.618 0.425 0.512 where 0.376 0.247 0.226 0.078 0.245 where(no OP) 0.385 0.286 0.301 0.117 0.284 group(no Having) 0.333 0.427 0.512 0.376 0.415 group 0.333 0.420 0.488 0.365 0.404 order 0.429 0.297 0.706 0.747 0.556 and/or 1.000 0.916 0.908 0.895 0.932 IUEN 0.000 0.000 0.057 0.125 0.074 keywords 0.835 0.808 0.703 0.618 0.757 ---------------------- PARTIAL MATCHING RECALL ---------------------- select 0.664 0.393 0.615 0.418 0.500 select(no AGG) 0.676 0.409 0.615 0.418 0.510 where 0.380 0.253 0.228 0.061 0.237 where(no OP) 0.389 0.292 0.304 0.092 0.275 group(no Having) 0.500 0.427 0.538 0.405 0.442 group 0.500 0.420 0.513 0.392 0.431 order 0.545 0.293 0.610 0.728 0.544 and/or 0.992 0.953 0.969 0.942 0.964 IUEN 0.000 0.000 0.048 0.056 0.051 keywords 0.913 0.804 0.667 0.600 0.755 ---------------------- PARTIAL MATCHING F1 -------------------------- select 0.664 0.394 0.617 0.421 0.501 select(no AGG) 0.676 0.410 0.617 0.421 0.511 where 0.378 0.250 0.227 0.069 0.241 where(no OP) 0.387 0.289 0.303 0.103 0.280 group(no Having) 0.400 0.427 0.525 0.390 0.428 group 0.400 0.420 0.500 0.378 0.417 order 0.480 0.295 0.655 0.737 0.550 and/or 0.996 0.934 0.937 0.918 0.947 IUEN 1.000 1.000 0.052 0.077 0.061 keywords 0.873 0.806 0.684 0.609 0.756
The results on Dev set are close to our original results. You may submit the code as a PR so that we can merge it into ours. Again, please tell us if you meet any problem evaluating using Codalab. This is helpful for us to improve the usability of our evaluation method.
@tomsherborne Would you like to provide us with your updated code? This could be helpful for follow-up research. Thank you very much!
Hi, Do you have any plans to update the code yourself in the future? I've been trying to replicate the results from the paper but I wanted to use a more up to date Python and PyTorch version. Therefore, I've spent some time modifying the model to run with PY3.7 and Pytorch 1.2 (not >1.3.1 as I'm testing using a K40 GPU currently). If you didn't have a plan to make these changes, I have a version with minimal changes from the original code that I can submit as a PR if you'd like to use.