A conditional total for routed expert weight traffic

Read the equation

Multiplying the dossier’s 44,050,944 bytes per expert by 464 expert evaluations gives 20,439,638,016 bytes, about 19.04 GiB. This is the stated one-load-per-evaluation accounting assumption. Reuse, resident weights, batches, scale tensors and other model components affect actual storage traffic, so this is not an observed disk-throughput measurement.

Terms & definitions

The recovered passage does not supply a separate symbol table. Read the source context for the notation used in this version; familiar letters may carry different meanings in other builds.

How it works

Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.

Assumptions & limits

Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.

Original context

ROUTED EXPERT TRAFFIC (streamed, per token)
      8 experts x 58 MoE layers                 = 464 expert reads per token
      464 x 44,050,944 B = 20,439,638,016 B     = 19.04 GiB per token

  Against the measured drive rates from PART 1:
  ---------------------------------------------------------------------------
      at   507 MB/s  single stream                    40.3 s per token

464 x 44,050,944 B = 20,439,638,016 B     = 19.04 GiB per token

Related builds