Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Is there a problem with the 'original' weight type and base model combination?
#3
by
jojo0217
- opened
I submitted "jojo0217/ChatSKKU5.8B" model with chosing the weight type as original but accidentally put base model section.
the results show that most of the accuracy is similar to and even completely identical to the base model.
Thank you for your report.
When evaluating process, we only look at the section of the model name (“jojo0217/ChatSKKU5.8B”) and ignore the base model (EleutherAI/polyglot-ko-5.8b). So, any similar scores probably come from how “jojo0217/ChatSKKU5.8B” performs, not the base model. If you notice anything unusual, please let me know. I'll double-check it.
Thank you for your check
jojo0217
changed discussion status to
closed