Training on new data in scierc format

[Kehindeajayi01]

Hi dwadden, Thanks for the great work. I am trying to train your model on a new data that's in scierc format. I have used the brat_input.py to convert my .ann and .txt to scierc format. However when I ran the training script on my data, I got the error below:

[dwadden]

Hi,

Can you please attach a minimal example (ideally a single line from a jsonl file) that reproduces this error, along with the model training command you use to start training?

Dave

[Kehindeajayi01]

Hi,

Can you please attach a minimal example (ideally a single line from a jsonl file) that reproduces this error, along with the model training command you use to start training?

Dave

This is the training command I used: bash scripts/train.sh scierc_lightweight

[dwadden]

Sorry for the slow response. Could you include the example as a file attachment rather than as a screenshot?

[Kehindeajayi01]

Sorry for the slow response. Could you include the example as a file attachment rather than as a screenshot? Please find the attached for the requested file.

dygie_example.txt

Thanks

[dwadden]

In your example, one of your relations has value [109, 109, null, null, "has_unit"]. The code is breaking due to the presence of null values. See data.md on the required format for input data.

[dwadden]

I'll close this. Feel free to re-open if this doesn't solve your problem.

[Kehindeajayi01]

In your example, one of your relations has value [109, 109, null, null, "has_unit"]. The code is breaking due to the presence of null values. See data.md on the required format for input data.

The null were introduced by your brat_input.py script when I used it to convert the brat format to your format

[dwadden]

Ah I see, sorry for closing prematurely.

brat_to_input.py was added by a contributor; more info here. I unfortunately can't offer support, but maybe @serenalotreck can help?