Generative QA with LFQA

[Chetan-Yeola]

Can we train a reader/generator on our own dataset?

[mayankjobanputra]

Yes you can. Haystack doesn't support training these models but It is like fine-tuning any other Seq2Seq model. Here is an example

[Chetan-Yeola]

Hi, @mayankjobanputra Thanks for your reply. but the link is dead.

[mayankjobanputra]

@Chetan-Yeola I will talk with the team about this issue and get back to you!

[mayankjobanputra]

@Chetan-Yeola the link should be working now. Please let me know if you aren't able to access it.

[Chetan-Yeola]

@mayankjobanputra Yes. the link is working now. Thank you..!

[OwaisAzad]

Yes @mayankjobanputra , you are correct. Haystack is an open-source library for building and evaluating information retrieval systems, and it does not currently support training models for the LFQA task. However, it is possible to fine-tune an existing Seq2Seq (Sequence-to-Sequence) model for the task of generative QA, including LFQA.