Benchmarks

Benchmarks

Reproducible measurements of NeuronScope on real hardware. Every row links to the .nsz artifact and the exact commit that produced it.

8 / 8 rows
ModelParamsDeviceProbesCommit
distilbert-base-uncased66MA100 40GBactivations1,840620419c1a04f
bert-base-uncased110MA100 40GBactivations1,210980729c1a04f
gpt2124MA100 40GBactivations+attention7801,4401689c1a04f
bert-large-uncased340MA100 40GBactivations4202,8602149c1a04f
gpt2-medium355MA100 40GBactivations+attention2403,8205129c1a04f
distilbert-base-uncased66MM3 Maxactivations240610419c1a04f
llama-3.1-8b8BH100 80GBactivations (stride=2)11018,4001,2809c1a04f
distilbert-base-uncased66MCPU (16c)activations72580419c1a04f

Methodology: batch size 32, tokens per row 128, ring buffer 1 GiB, warm cache. Numbers are the median of 5 runs after 100 rows of warm-up. Reproduce with bench/run.py at the commit in the last column.