hec-ovi/vllm-awq4-qwen

vLLM Qwen 3.6-27B (AWQ-INT4) + DFlash speculative decoding on AMD Strix Halo (gfx1151 iGPU, 128 GB UMA, ROCm 7.13). 24.8 t/s single-stream, vision, tool calling, 256K context, OpenAI-compatible, Docker. Matches DGX Spark FP8+DFlash+MTP at a third of the cost. No CUDA.

Python

awqdflashdockergfx1151llm-inferencemultimodal-llm

OPEN ON GITHUB

// 01 · GITHUB MOMENTUM

12.9

0 stars 24h | 0 7d

// 02 · MENTIONS

0 in 24h | 0 sources

// 03 · CROSS-SIGNAL

0.00

0/5 channels firing

// 04 · PACKAGE ADOPTION

// STARS16—/ 7d

// FORKS0—/ 7d

// CONTRIBUTORS0—/ 30d

// CROSS-SIGNAL BREAKDOWN · PER-CHANNEL COMPONENTS0.00/5.0 · 0/5 FIRING

How is this scored?

5.0: strong signal across ≥4 of 5 channels
4.0: strong signal across ≥3 of 5 channels
3.0: strong signal across ≥2 of 5 channels
2.0+: active on at least 1 channel
<1.0: low or no cross-channel activity

Each channel contributes 0-1. Per-channel tiers: GitHub (breakout 1.0 / hot 0.7 / rising 0.4), HN (front-page 1.0 / ≥3 mentions 0.7 / 1-2 mentions 0.4), Bluesky (≥5 mentions 1.0 / 2-4 0.7 / 1 0.4), dev.to (≥3 articles 1.0 / 2 0.7 / 1 0.4), Reddit (corpus-normalized 48h velocity).

// MENTIONS · EVIDENCE FEED0 / 0

// last scanreddit 43mhn 1hbluesky 4mdevto 11hph 2hnpm 10h

No mentions on this channel in the last 7 days.

// QUIET HERE DOESN'T MEAN THE REPO IS DEAD — CHECK OTHER TABS

// RELATED REPOS6 REPOS