allanj / deductive-mwp

Hi there 👋

deductive-mwp's Issues

Do you have check points for MAWPS?

I would like to ask you if you can provide the check points of the model with MAWPS as well.

Do you have the dataset of SVAMP which that used for your model?

I cannot find the specific dataset that you used to train/test SVAMP for your model.
Do you have one?

A question about dataset evaluation.

5-fold cross validation is used on the MAWPS dataset. For each of the five experiments, should I choose the best result in the validation set or use other results？

How to process datasets into structures

Cannot run your training code

Hi,

Thanks for your work.

I just downloaded your latest code and tried to reproduce your results. However, I encountered a bug saying
"universal_main.py: error: unrecognized arguments: --add_new_token=0 --diff_param_for_height=0 --filtered_steps -1".

Even after I manually removed these three parameters, there are still bugs saying
"FileNotFoundError: [Errno 2] No such file or directory: 'data/mawps-single/mawps_train_nodup.json'".

Do you know how to fix the issue?

Thanks!

There is a relative reasearch

《A Bottom-Up DAG Structure Extraction Model for Math Word Problems》AAAI 2021，framework and perspective are the same as you. They do not have Rationalizer and the other part of the model is the same as u.

U did't read that? But I saw u refer a relative paper in AAAI 2021<Hms: A hierarchical solver with dependency-enhanced understanding for math word problem.>

I wonder to know where you put the constraint

I read your paper "Learning to Reason Deductively: Math Word Problem Solving as Complex Relation Extraction".

In Chapter 3.3, The paper discusses the two constraint

1. there does not exist any expression that involve operations applied to the same quantity
2. intermediate results would not be negative

but I can't find these constraints in your code yet.
Can you tell me which part of the code I should refer to?

Is there any way to use your pre-trained model?

I would like to use your pre-trained model.
Would there be any way to use it?

I don't know how to run your code

Hello,

Thank you for the code.But I ran into some problems,I don't know how to run your code.

I wanted to reproduce your training results, but I packed up the entire file and uploaded it to Google Colab and didn't know how to run it, like you said bash scripts/run_math23k.sh, but error name 'bash' is not defined. I am a newcomer and do not understand the code, can you help me solve and successfully reproduce?

About the released checkpoints

It seems that the released checkpoints for MathQA and SVAMP have significantly worse performance than reported in the paper.