Debugger

Using TokenPrint

Debugger mode is a tiled engineer dashboard for deep model inspection. It is organized into panels that can be rearranged. All data comes from real model calls to /debug/* and /ablate/analyze.


Panels

PanelWhat it does
BreakpointsSet layer/head breakpoints. Playback pauses when the specified component executes.
Flame graphPer-op timing breakdown showing host-side execution times. PROXY label on MPS/CUDA (no device sync).
Anomaly sentinelsAutomatic alerts for NaN, Inf, near-zero activations, and attention collapse patterns.
Watch expressionsMonitor specific tensor values across layers or steps.
Layer tablePer-layer summary: residual delta, norm magnitude, top attention weight.
Head gridAll 14 × all layers attention heatmaps in a compact grid view.
AblationZero out a selected head or layer and compare before/after logit distributions. Uses POST /ablate/analyze.
Induction-head labIdentifies candidate induction heads by their characteristic copy-suppression attention pattern.
Trace replayReplay any saved trace. Branch from a replay to run a modified forward pass and compare the two traces side by side.
Console REPLSend arbitrary commands to the debug API. Useful for extracting tensor values not exposed by the standard panels.

Activation patching

Warning

The “Activation Patching” panel title is a known inaccuracy — the panel shows a logit-lens trajectory (a correlational measurement), not causal activation patching. Tracked in issue #75.