---
title: '50424'
pageLayout: default
slug: thoughts/inference/vllm-prs/50424
permalink: https://aarnphm.xyz/thoughts/inference/vllm-prs/50424.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
# [#50424](https://github.com/aarnphm/aarnphm.github.io/issues/50424) Support quantized DSpark Markov heads

[vllm-project/vllm#50424](https://github.com/vllm-project/vllm/pull/50424) · askliar · MERGED · created 2026-07-30T11:20:43Z · merged 2026-08-03T12:35:40Z

## Summary

Add quantization configuration support to the DSpark Markov head.

`DSparkMarkovHead` now accepts and forwards `quant_config` to its `ParallelLMHead`-based `markov_w2` projection. `Qwen3DSparkModel` passes its model quantization configuration when constructing the Markov head.

This allows quantized `markov_w2` weights, including W4A16 configurations with `weight_scale_2`, to load through the normal quantization dispatch path while preserving unquantized behavior.

## Testing

Attempted DSpark quantization and loading with this branch.

---

## comments

