As LLM becomes larger and larger, some highly quantized types have emerged in recent years, such as `FP8`, `Q4`. Should `tensor-type` add these definitions, or provide `custom(string)` enumeration
As LLM becomes larger and larger, some highly quantized types have emerged in recent years, such as
FP8,Q4.Should
tensor-typeadd these definitions, or providecustom(string)enumeration