h3-blackwell-runtime/benchmarks/gb10-cute-p2-ring-full-block0-attn-out.json
2026-08-25 20:30:22 +07:00

205 lines
4.2 KiB
JSON

{
"device": "NVIDIA GB10",
"cutlass_dsl": "4.6.2",
"metadata": {
"width": 1344,
"height": 768,
"frames": 124,
"steps": 12,
"sampler_step": 1,
"seed": 440420,
"text_tokens": 100,
"hidden_shape": [
37810,
5376
],
"segments": [
[
0,
100,
1
],
[
100,
514,
2
],
[
514,
37810,
0
]
]
},
"block_index": 0,
"capacity": 2048,
"all_equal": true,
"projection_reference_total_ms": 25.982784271240234,
"projection_ring_total_ms": 14.904607772827148,
"results": [
{
"name": "attn_out_proj",
"mnk": [
37810,
5376,
7168
],
"capacity": 2048,
"chunk_count": 19,
"final_chunk_rows": 946,
"ring_bytes": 8257536,
"parity": {
"all_chunks_equal": true,
"max_abs": 0.0,
"mean_abs_max": 0.0,
"reference_checksum": -55638368.0,
"candidate_checksum": -55638368.0
},
"timing": {
"warmup": 1,
"iterations": 1,
"ring_complete_ms": 14.904607772827148,
"reference_complete_ms": 25.982784271240234,
"ring_vs_reference": 0.5736339730659553,
"improvement_percent": 42.63660269340447
},
"chunks": [
{
"start": 0,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 2048,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 4096,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 6144,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 8192,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 10240,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 12288,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 14336,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 16384,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 18432,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 20480,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 22528,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 24576,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 26624,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 28672,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 30720,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 32768,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 34816,
"rows": 2048,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
},
{
"start": 36864,
"rows": 946,
"equal": true,
"max_abs": 0.0,
"mean_abs": 0.0
}
],
"output_bytes": 418381824
}
],
"projection_total_improvement_percent": 42.63660269340447
}