Model size

#1
by depasquale - opened
MLX Community org

Was this model quantized correctly? It appears to be over 60 GB in size, while the previous 4-bit version was around 17 GB.

MLX Community org

@awni , the total size of the safetensors files is no smaller than the original unquantized model. Something seems to have gone wrong with the quantization here.

MLX Community org

It's been updated.

Sign up or log in to comment