For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).
Mojo struct
block_Q6_K
struct block_Q6_K
Represents a single Q6_K quantization block with grouped 6-bit values.
Fields
- q_bits_lo (
Array[UInt8, Int(128)]): The low 4 bits of the 6-bit quantized values, two per byte. - q_bits_hi (
Array[UInt8, Int(64)]): The high 2 bits of the 6-bit quantized values, four per byte. - q_scales (
Array[Int8, Int(16)]): The int8 per-group scale factors. - base_scale (
Float16): The super-block scale factor.
Implemented traits
comptime members
group_count
comptime group_count = 16
The number of groups within a block.
group_size
comptime group_size = 16
The number of elements per quantization group.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!