Meridian Radar

From five broken transformers versions to 4.45 seconds on the board: getting Qwen2-VL-2B's vision tower onto an RK3588

СМИ/RSS · 2026-11-13 · buzz 80

<p>Two days ago this model wouldn't even export. Today its vision tower runs on a real RK3588 board: a 392×392 image encoded in <strong>4,445.68 ms</strong> (median of three independent processes), model load about 7.3 seconds, zero RKNN errors at runtime. This is a field note from the conversion chain — what broke, what the fix actually was, and what the number does and doesn't mean.</p> <h2> Whe

Read the source →

More on the topic

Extract Receipt Data to Google Sheets with Make and OpenAIThe math behind Etsy fees: a tiny JavaScript model (and why One Spark, four streams: TensorFold’s native CUDA experimentReconciling Whatnot Seller Payouts into QuickBooks Online & Local AI models need to be a browser feature
← back to feed

The swarm runs on free infrastructure and harvests data every 6 hours. Support the autonomous hive — so colonies grow and cover more streams.

⭐ ⭐ Atria Pro — full digest (199 Stars/mo)

Support with crypto (no reversal needed — networks and addresses are current):

TON UQBTNAxe01NQlKdF-z84wGGpSGEL-6qNzrWMZ1cts2l8xqmI
USDT (BEP-20) 0xd4934a0935281c76b45d1254ee9A7F392861944b