vllm/csrc/quantization/machete

Machete (Mixed Precision Cutlass-Based GEMM)

Machete is a spiritual successor to the Marlin kernel but optimized for Hopper architectures and based on Cutlass. Being based on Cutlass, new type pairs and epilogues are easier to add compared to Marlin.

Overview

Machete effectively performs

scale_type = w_s.dtype
compute_type = a.dtype
out = (w_q.to(scale_type) * w_s - w_z.to(scale_type)) @ a

Where w_q is a quantized weight matrix, w_s is the quantization scales, and w_z is the quantization zeropoints.

NOTE: w_z is added after the scales so we can use FMA operations, but this means they must have the scales pre-applied if the supplied zeropoints assume that they will be subtracted before the scales are applied.

API

The main optimization within Machete is prepacking the weight matrix to more closely match the tensor core layouts, allowing for wider shared memory loads when loading the weight matrix. This means that the weight matrix must be prepacked before calling machete_gemm. The flow looks something like:

from vllm import _custom_ops as ops

...
W_q_packed = ops.machete_prepack_B(w_q, wtype)
output = ops.machete_gemm(
    a,
    b_q=W_q_packed,
    b_type=wtype,
    b_scales=w_s,
    b_group_size=group_size
)

Code Generation

Since Machete is based on Cutlass, we can generate multiple type pairs and different tile shapes using the same kernel template. We generate multiple instantiations of this template using generate.py.

New type pairs (TypeConfigs) can be appended to impl_configs (in generate()), and these will get automatically generated (assuming they can be supported without issues). For each TypeConfig, you must also provide an ImplConfig, which bundles a TypeConfig with a list of ScheduleConfigs, Specializations, and a default heuristic. The ScheduleConfigs (which contain info on tile shapes, tile scheduler, etc.) can perform differently for different problem shapes, and there is almost never one ScheduleConfig that works well for all problem shapes, so it is generally beneficial to generate different ScheduleConfigs for different potential problem shapes. This is where the heuristic comes in. For each TypeConfig, a default heuristic should be provided. This maps different problem shapes to different ScheduleConfigs and is used when the user does not provide the schedule parameter to machete_gemm. The Specializations define what feature combinations to generate, i.e., with_zeropoints, with_scales, etc. We can reduce compile times and the final binary size by limiting the set of feature combinations we generate.