
What happened
On a Dragonwing RB3 Gen 2, 6 of 9 InternVL3.5-2B graphs profiled via the manual AI Hub chain. Parts 2–4 totaled 174,113 µs for a 128-token prompt and 62,815 µs per decoded token; Part 1 and the vision encoder failed.
Why it matters
The failure traces to a single unquantized layer-2 tensor in Part 1, blocking a complete end-to-end measurement — so the 6-graph subtotal is not presented as the model's full performance.
Summaries like this, every morning.
This profiling attempt follows two earlier LLM runs on the same QCS6490 (Dragonwing RB3 Gen 2) using a manual AI Hub export chain, where Qwen3-0.6B reached 1,555 tok/s prefill and Llama 3.2 1B reached 1,024. InternVL3.5-2B was the natural next target: it pairs an InternViT vision encoder with a Qwen3-series text backbone, carries an Apache-2.0 license, and had no public measurements on this chip. The export script itself was not the obstacle this time; the blocker sat inside Qualcomm's own quantized asset. The linker error names a specific tensor — layer 2's down_proj Conv bias — as having a floating-point type unsupported by the target device, and the same class of problem appears in the vision encoder while corresponding tensors in Parts 2–4 compose and run normally. A single bounded experiment recompiling Part 1 at -O1 instead of -O0 left the error unchanged, pointing to the shipped asset rather than compile options. The official Genie pipeline presumably handles these tensors somewhere locally, but the manual chain bypasses it and collides head-on. An issue with job numbers and error strings is being filed with the vendor, mirroring the process used when the export script itself had a bug. Reproduction also surfaced three side traps: the model template pins transformers 5.7.0, incompatible with the 4.45 environment used for Qwen3 and Llama; aimet-onnx has no aarch64 wheel, requiring a no-op patch to an eager QuantSim initializer that was then verified by sha256 byte-match against the vendor's original ONNX; and the 9.39 GB asset download silently froze at 70%, resolved by resuming with curl -C -. The series now stands at two LLMs measured end to end and one VLM measured for the 6 graphs the vendor asset permits, with the blocker isolated to a specific tensor in a specific layer — an honest first public record either way.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
BOOKERS now supports MCP, so users can connect their account to Claude or ChatGPT (web version, confirmed work…

IMPACTLAKE began offering a ChatGPT plugin for impactlake(TM) and added Model Context Protocol (MCP) support…

Google DeepMind introduced Gemini 4 Hydrogen and Gemini 4 Carbon, with Hydrogen being integrated into Google S…

The author installed the open-source agent platform OpenClaw on an EPSON Endeavor NY2100S (Celeron P4600, 4GB…

A freelance web and SEO tool developer reports running Claude Code with one COO instance, 15 sub-agent departm…

A developer using Claude Code on a personal web app discovered via --debug logs that his Stop hook passed ever…
