The Size of APIGen-MT Dataset Used for Training xLAM-2-fc-r Model

[ikkvix]

Thank you for APIGen-MT's valuable contribution to the open-source community!

I noticed that the paper mentions 5k open-source data, but it doesn't specify the actual amount of data used during training. Could you please clarify how much data was actually used when training the xLAM-2-fc-r model?

I'm curious about this because understanding the scale of training data would help the community better understand the model's capabilities and potentially reproduce similar results.

Thanks in advance!

[Sakura171]

Hello, I’m also working on reproducing xLAM and trying to train with a 5k dataset. However, so far my training has shown minimal results for multi-turn scenarios. I was wondering if you managed to reproduce it successfully? How did you process the data specifically?