SAM hq vision encoder window partition attention
SAM-HQ's windowed attention mechanism that fuses window partitioning, multi-head attention with 2D relative position embeddings, window unpartitioning, residual connections, layer normalization, and MLP block. Processes image patches in 14x14 windows for efficient local attention computation.
Reported evidence · last observed 2026-07-30. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencymlp_dim = 3072 · head_dim = 64 · hidden_size = 768 · window_size = 14 · qkv_out_size = 2304 · rel_pos_size = 27 · num_attention_heads = 1216 cases
Not measured on H100 for this workload. Challenges →
12.24ms88% SOL1.00×Reported · license unknown · no source2026-07-30
Measured exactly what you asked. Reported by source; not independently reproduced.
22.67ms84% SOL1.19×Reported · license unknown · no source2026-07-11
Measured exactly what you asked. Reported by source; not independently reproduced.
32.67ms84% SOL1.19×Reported · license unknown · no source2026-07-18
Measured exactly what you asked. Reported by source; not independently reproduced.
42.72ms84% SOL1.21×Reported · license unknown · no source2026-08-03
Measured exactly what you asked. Reported by source; not independently reproduced.
53.02ms82% SOL1.35×Reported · license unknown · no source2026-08-24
Measured exactly what you asked. Reported by source; not independently reproduced.
63.28ms80% SOL1.47×Reported · license unknown · no source2026-08-22
Measured exactly what you asked. Reported by source; not independently reproduced.
73.34ms80% SOL1.49×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
83.34ms80% SOL1.49×Reported · license unknown · no source2026-07-12
Measured exactly what you asked. Reported by source; not independently reproduced.
93.44ms79% SOL1.54×Reported · license unknown · no source2026-08-17
Measured exactly what you asked. Reported by source; not independently reproduced.
103.44ms79% SOL1.54×Reported · license unknown · no source2026-08-12
Measured exactly what you asked. Reported by source; not independently reproduced.
113.93ms76% SOL1.75×Reported · license unknown · no source2026-07-19
Measured exactly what you asked. Reported by source; not independently reproduced.
124.68ms71% SOL2.09×Reported · license unknown · no source2026-08-10
Measured exactly what you asked. Reported by source; not independently reproduced.
134.68ms71% SOL2.09×Reported · license unknown · no source2026-08-11
Measured exactly what you asked. Reported by source; not independently reproduced.
146.13ms64% SOL2.74×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.
156.54ms63% SOL2.92×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1612.1ms46% SOL5.39×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1741.1ms19% SOL18.35×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.