Does this model use GTE weights?

#14
by nauti16 - opened

Hi, thank you for sharing this great model!!!

I understand that arctic-embed-m-v2.0 builds on the GTE-multilingual-base.
To clarify whether this model supports commercial use, could you confirm:

  1. Does 'arctic-embed-m-v2.0' reuse the pre-trained weights from 'GTE-multilingual-base', or
  2. Did you train the model entirely from scratch using your own data without pre-trained weights from GTE-multilingual-base?

Because GTE-multilingual-base was trained on MS MARCO, which is restricted to non-commercial use.

Thanks in advance!

Snowflake org

We trained arctic embed 2.0 m based on ‘Alibaba-NLP/gte-multilingual-mlm-base’, which represents weights before fine tuning on MS MARCO.

Thank you for the clarification!!!

pxyu changed discussion status to closed

Sign up or log in to comment