vllm-project/vllm

[Feature]: Refactor Int8ScaledMMLinearLayerConfig to use QuantKey

Open

#32.268 aberto em 13 de jan. de 2026

Ver no GitHub
 (7 comments) (0 reactions) (1 assignee)Python (16.816 forks)batch import
feature requestgood first issuehelp wanted

Métricas do repositório

Stars
 (80.034 stars)
Métricas de merge de PR
 (Mesclagem média 3d 17h) (993 fundiu PRs em 30d)

Description

🚀 The feature, motivation and pitch

Replace boolean configuration fields in ScaledMMLinearLayerConfig with QuantKey objects to provide a more structured, type-safe quantization configuration API.

Ideally we should change this:

@dataclass
class ScaledMMLinearLayerConfig(ScaledMMLinearLayerConfig):
    is_static_input_scheme: bool
    is_channelwise: bool
    input_symmetric: bool

to this:

@dataclass
class ScaledMMLinearLayerConfig(ScaledMMLinearLayerConfig):
    weight_quant_key: QuantKey
    activation_quant_key: QuantKey
    input_symmetric: bool

A parallel work found here #27814 , has split the configuration into Int8 and Fp8 config classes and uses Quantkey for the FP8 config class.

Alternatives

No response

Additional context

No response

Before submitting a new issue...

  • Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

Guia do colaborador