Can you release your training code ?

[HuyTT1611]

Can you release your training code ?

[youcan9021-wq]

waiting training code

[Carkham]

Hi all, we are planning to release our training code in one week. Stay tuned!

[youcan9021-wq]

I have numerous unlabeled table datasets, so I would like to utilize this approach.

[youcan9021-wq]

@Carkham
Any update about training pipelines and code release.

[Carkham]

Thanks for your attention and patience; we have released our training code. Enjoy it.

[youcan9021-wq]

@Carkham
thank you for releasing training code.

could you please provide the sample dataset prepared for GRPO training. I don't understand how qa_pairs generate for OTSL training.
{
"messages": [ // Prompt and label for table recognition. Should be replaced by your owns.
{"role": "user", "content": "You are an AI specialized in recognizing and extracting table from the image. Your mission is to analyze the table in the image and generate the result in OTSL format using specified tags.\n"},
{"role": "assistant", "content": "

"}
],
"qa_pairs": [ // A list of qa pairs generated for this image.
{"question": "What is the unit listed for all values under Isc current?", "answer": "mA"},
...
],
"images": ["/path/to/img"]
}

[Carkham]

@youcan9021-wq Hi. We have included a sample dataset. Please refer to https://github.com/opendatalab/TRivia/tree/main/training#data-preparation