[theanonymousone]

But the huggingface link mentions BF16, F16, and I32?

[kouteiheika]

Not every weight is quantized. For example, those weights which don't take much space or are highly important are left in higher precision. State-of-art quantization of weights is never done uniformly (i.e. to all weights and in the same way).

[zackangelo]

I don't believe safetensors has a native int4 dtype, so they packed 4 int4s into a bf16 in this checkpoint.