could we create 139B quantization with REAP for this model?
#2
by
Slach
- opened
177B doesn't fit to my GB10 ;)
On the todo list!
It’s this a REAP version or the full the Params?
It’s this a REAP version or the full the Params?
This repo's model is the full params (though quantized)
Why the model card says 177B params and not 358B as the original model - is it a typo?
HuggingFace automatically pulled that number from the safetensors. I've seen that issue on other model cards as well. Not sure how to correct it in all honesty