Vision patch merger spatial shuffle MLP
Vision patch merger that performs spatial shuffling to merge patches (spatial_merge_size=2), applies layer normalization, and processes through a 2-layer MLP with GELU activation. Converts fine-grained vision tokens into coarser representations by merging 2x2 spatial patches.
Reported evidence · last observed 2026-07-10. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 15 cases · mean latencythree = 3 · hidden_size = 1536 · out_hidden_size = 3584 · spatial_merge_size = 2 · hidden_size_expanded = 614415 cases
Not measured on H100 for this workload. Challenges →
1144.2µs73% SOL1.00×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
2147.5µs72% SOL1.02×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
3155.2µs70% SOL1.08×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
4156.2µs70% SOL1.08×Reported · license unknown · no source2026-07-28
Measured exactly what you asked. Reported by source; not independently reproduced.
5162.5µs67% SOL1.13×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
6164.5µs67% SOL1.14×Reported · license unknown · no source2026-07-19
Measured exactly what you asked. Reported by source; not independently reproduced.
7166.9µs66% SOL1.16×Reported · license unknown · no source2026-08-11
Measured exactly what you asked. Reported by source; not independently reproduced.
8167.0µs66% SOL1.16×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.
9168.7µs66% SOL1.17×Reported · license unknown · no source2026-07-12
Measured exactly what you asked. Reported by source; not independently reproduced.
10169.4µs66% SOL1.17×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
11175.5µs64% SOL1.22×Reported · license unknown · no source2026-08-06
Measured exactly what you asked. Reported by source; not independently reproduced.
12213.0µs55% SOL1.48×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
13230.8µs53% SOL1.60×Reported · license unknown · no source2026-07-17
Measured exactly what you asked. Reported by source; not independently reproduced.
14262.4µs49% SOL1.82×Reported · license unknown · no source2026-07-23
Measured exactly what you asked. Reported by source; not independently reproduced.
15267.5µs48% SOL1.85×Reported · license unknown · no source2026-07-24
Measured exactly what you asked. Reported by source; not independently reproduced.