Can I run this on my own table graph and use that information to generate queries for my database given text?

#3
by Prudhviraj - opened

Hi, I would like to use this model for some text to sql research and I want to see how well this model can adapt to just switching the graph component of it and doing minimal finetuning.

@Prudhviraj Of course, we also use this for our other table-based research with different contents

Thank you for asking

Hi @patrickNLP ,
I was trying something similar, to run your model and check the accuracy on my dataset, however, while I am running it on google collab (free version with 12 GB RAM and 4 GB T4 GPU), it was getting terminated saying: "Your session crashed after using all available RAM. If you are interested in access to high-RAM runtimes, you may want to check out Colab Pro".

Is this because, the model is too heavy or my code is wrong? Could you kindly guide?
Thank you @Prudhviraj for asking the question.

@AanVar ,

Hi, thanks for your asking. It seems that your GPU is too small to accommodate such a model. I'm not aware of what is the goal of yours to run this model. If you are seeking methods for products, well, most academic models are not a good choice except coda-llama2 and ChatGPT. If you are doing this for research, and compare results against Graphix, you can pull a request at the GitHub site of this model and request them to generate the predicted data via inference. Thanks.

Sign up or log in to comment