TokenPrint turns Qwen inference into a DevTools-style visual debugger
Rich-Fruit-326 · reddit · 2026-07-21
**TokenPrint** is an open-source project that aims to make transformer inference debuggable, like Chrome DevTools for LLMs. The author says current workflows rely on logs, tensors, or notebooks, but lack an interactive way to step through a forward pass. The project already supports: - live **Qwen** inference - token-by-token execution - tensor inspector and tensor grid - architecture explorer - replay system in progress - KV cache, GQA, and RoPE visualization - interactive operation inspection The long-term goal is a visual debugger where users can inspect activations, trace tensors, compare runs, replay execution, and answer questions like “why did the model generate this token?” The screenshots show a dark-themed debugger UI built around a **Qwen2.5-0.5B-Instruct** model.
More from coding & agent
- uv-scripts/ocr returns to the top of Hugging Face datasets with a JSON model picker — vanstriendaniel · 2026-07-21
- Sonar CEO says a guide-verify-solve loop cuts coding-agent issues by 92% — alex_verem · 2026-07-21
- A creator built an Awwwards-style landing page with ChatGPT 5.6 Sol in one conversation — paw_lean · 2026-07-21
- OpenAI’s Build Week buildathon drew 40 people for 11 hours with Codex — paw_lean · 2026-07-21
- Workshop to cover loop and graph engineering for AI-native software engineering — Al_Grigor · 2026-07-21
- Production agents may need a new PaaS layer built around audit and recovery — percoAi · 2026-07-21