Usable model

[ghost]

This might be a no brainer or too simple but I'm curious.

To use this we need a model in F16 or F32 format that others can then quantize to whatever is the SOTA, ie, the best currently.

Please provide information on whether that can be done with this model. The use-case is avoiding doing it in a way which involves the tedious process of setting up Docker and figuring out how to select, build, test a image and then do the simple things the convoluted way.

Optional - nothing else seems important really. A FP{16,32} model can be easily converted to another form.

Reason - it's unreasonable to assume everyone will use the safetensors model which not even the official site cares to set up.

Thanks.