bitnet-5B-v0 / README.md
Ttimofeyka's picture
Update README.md
f8d4e33 verified
|
raw
history blame
310 Bytes
metadata
license: mit

This model is my starting point zero for trying to finetune model based on bitnet architecture. I just added new layers with random weights to the finished model. Maybe it can be broken. It is not recommended for use: the results show an improvement in test results at the margin of error.