h3-blackwell-runtime/benchmarks/gb10-sage2-block24-mainloop-capture.json
2026-08-25 20:30:22 +07:00

72 lines
1 KiB
JSON

{
"metadata": {
"width": 1344,
"height": 768,
"frames": 124,
"steps": 12,
"sampler_step": 1,
"seed": 440420,
"text_tokens": 100,
"block_index": 24,
"attention": "sage2",
"hidden_shape": [
37810,
5376
],
"h_msa_shape": [
37810,
5376
],
"segments": [
[
0,
100,
1
],
[
100,
514,
2
],
[
514,
37810,
0
]
]
},
"capture": "one unchanged prequantized Sage2 mainloop",
"q_shape": [
1,
37810,
56,
128
],
"k_shape": [
1,
37810,
56,
128
],
"v_shape": [
1,
37810,
56,
128
],
"checksum": 9190064.0,
"scheduler": {
"cta_q": 128,
"cta_k": 64,
"warp_q": 32,
"warp_k": 64,
"warps_per_cta": 4,
"threads_per_cta": 128,
"dynamic_shared_memory_bytes": 32768,
"q_ctas_per_head": 296,
"heads": 56,
"grid_ctas": 16576,
"k_iterations_per_cta": 591,
"explicit_pipeline_stages": 2
}
}