Training loss?
Hey, do you have the training checkpoints/loss/evaluations/other-ablations-tried to share?
Small note, you don't state which languages the model is expected to perform well at which people would look for.
Likely won't be answered by the authors due to connectivity isssue of PRC. https://news.yahoo.com/open-source-ai-platform-hugging-202603150.html?guccounter=1#:~:text=Chinese%20users%20have%20complained%20about%20having%20connectivity%20issues
Oh god, HF? I knew about arXiv... Guess I need to devise a new way...
Hey, do you have the training checkpoints/loss/evaluations/other-ablations-tried to share?
Currently, no we don't. But we have published a technical-ish report: https://arxiv.org/abs/2309.16609. I wish there is something useful to you.
Small note, you don't state which languages the model is expected to perform well at which people would look for.
Please look at the tags: Chinese and English.
It can work for other languages, but not that well (from GitHub README):
Likely won't be answered by the authors
Just haven't got the time to check HF. GitHub should be more responsive.