Training Data

[tcondello]

I did some hunting but was not able to find any repository or reference to where the used to train this model is being shared is it on Huggingface? My assumption is that these being open foundation models the data would also be open.

[mrutkows]

@tcondello The last reference i found (canonical, from the IBM Granite home page on ibm.com) was that the data is described in this research paper:
https://www.ibm.com/downloads/documents/us-en/10a99803c92fdb35

In addition, i see in the "model card" for this model in HF has a section with links; for example:

https://huggingface.co/ibm-granite/granite-3.1-8b-instruct

Training Data: Overall, our SFT data is largely comprised of three key sources: 
(1) publicly available datasets with permissive license, 
(2) internal synthetic data targeting specific capabilities including long-context tasks, and 
(3) very small amounts of human-curated data. 
A detailed attribution of datasets can be found in the 
[Granite 3.0 Technical Report](https://github.com/ibm-granite/granite-3.0-language-models/blob/main/paper.pdf), 
[Granite 3.1 Technical Report (coming soon)](https://huggingface.co/collections/ibm-granite/granite-31-language-models-6751dbbf2f3389bec5c6f02d), and 
[Accompanying Author List](https://github.com/ibm-granite/granite-3.0-language-models/blob/main/author-ack.pdf).