AbstractPhil/beatrix-captured-interactive-inferences
Beatrix captured interactive inferences Byte-by-byte internals of real inference runs on mini-beatrix-2.5s, the 237.1M full-splat byte model (model). Nothing here is simulated or sampled from a proxy: each capture is one greedy generation, re-run through a single instrumented forward pass that walks the blocks by hand and records what every layer did at every byte. Each prompt is captured twice over the same byte sequence — once on the bare core and once with the library's top… See the full description on the dataset page: https://huggingface.co/datasets/AbstractPhil/beatrix-captured-interactive-inferences.
recapture: the anchor arrays are now stored as a mean plus a percentile-scaled deviation. The first version quantized the absolute value against one global scale, which put the byte-to-byte signal below a single step — the arrays looked right and carried almost nothing per byte. Adds the per-byte write and the board's effective width, and ships the rebuilt viewer
track every .u8 array with one wildcard rule, so new capture arrays are covered
the interactive viewer ships beside its data, and the card links the published page
byte-by-byte internals of two real inferences on mini-beatrix-2.5s, each captured twice over the same sequence (bare core and with the top arm mounted): splat address reads, blackboard mass, derived effective attention, bank dispatch, head address and per-byte predictions
initial commit