search

google-research / tapas

End-to-end neural table-text understanding models.
Apache License 2.0
1.15k stars 217 forks source link

It's not returning answer for my data #12

Closed vitallalaji closed 4 years ago

vitallalaji commented 4 years ago

Hi, I'm using tapas on open source data sets.when I tried with only fraction of data i.e. small data(200 to 300 rows) it's returning answer for queries but if the data is big(4000 rows) it's not returning answer but it's able to return table. It's prompting "can't convert too many rows and sometimes sequence too long". How can I resolve this issue ? I want to try on big data sets. will it work ? Note : my system is not configured with GPU support and cuda. And I'm performing SQA task.

mdmustafizurrahman commented 4 years ago

How long does it take to run the prediction in your machine without GPU? What is your test bacth size? I am also running in laptop the prediction using SQA, and it has been running almost like 50 minutes but still no prediction results.

vitallalaji commented 4 years ago

How long does it take to run the prediction in your machine without GPU? What is your test bacth size? I am also running in laptop the prediction using SQA, and it has been running almost like 50 minutes but still no prediction results.

Its taking around 10 minutes to give the results because of my RAM (24 gb)..and my length of the queries is 3

muelletm commented 4 years ago

The problem of long sequences is addressed in #14. Does that help?