perf(dsv41): prompt-chunk partials over a reduce-scatter, the same bits from 3/4 of the bytes - #49

perf(dsv41): prompt-chunk partials over a reduce-scatter, the same bits from 3/4 of the bytes#49

imD-5 wants to merge 1 commit into

Elevatormusic:deepseek-v41-tp2Elevatormusic/TensorFold:deepseek-v41-tp2from

imD-5:perf/dsv41-reduce-scatterimD-5/TensorFold-2x-spark:perf/dsv41-reduce-scatterCopy head branch name to clipboard