# MemVanta trained-model layer profile
prefill_ms=44068.6 decode_total_ms=8827.1 sampling_ms=0.354846
prefill_cycles_per_token=841919363.73 decode_cycles_per_token=1349113453.62
qkv_projection_ms=12365.82 attention_output_projection_ms=3773.46 ffn_gemm_ms=32269.69 output_head_ms=3205.31
prefill_qkv_projection_ms=10980.84 prefill_attention_output_projection_ms=3300.34 prefill_ffn_gemm_ms=28505.58 prefill_output_head_ms=375.37 prefill_kernel_ms=43162.13
decode_qkv_projection_ms=1384.98 decode_attention_output_projection_ms=473.12 decode_ffn_gemm_ms=3764.11 decode_output_head_ms=2829.94 decode_kernel_ms=8452.16
non_gemm_core_ms=1281.10 (attention softmax/KV + RMSNorm + RoPE + elementwise/residual)
layer,qkv_ms,o_proj_ms,ffn_ms,total_projection_ms,pct_kernel_ms,projection_ms_per_token
0,386.92,118.50,1009.58,1515.00,2.94,10.52
1,386.90,116.88,1005.91,1509.69,2.92,10.48
2,387.44,119.81,1003.97,1511.22,2.93,10.49
3,388.77,118.21,992.40,1499.38,2.90,10.41
4,383.43,117.47,992.53,1493.43,2.89,10.37
5,384.37,117.45,1022.66,1524.48,2.95,10.59
6,390.03,117.95,1004.96,1512.94,2.93,10.51
7,385.98,117.27,1019.35,1522.59,2.95,10.57
8,387.59,116.89,1000.27,1504.75,2.92,10.45
9,383.55,116.45,1022.05,1522.05,2.95,10.57
10,384.43,116.49,1000.89,1501.81,2.91,10.43
11,384.28,117.76,998.81,1500.86,2.91,10.42
12,387.91,121.97,1005.74,1515.62,2.94,10.53
13,385.00,117.93,1007.52,1510.45,2.93,10.49
14,385.95,118.48,997.12,1501.55,2.91,10.43
15,391.00,119.21,1012.38,1522.59,2.95,10.57
16,386.56,116.48,1014.21,1517.25,2.94,10.54
17,384.19,117.33,992.20,1493.73,2.89,10.37
18,383.43,119.48,1000.13,1503.05,2.91,10.44
19,389.10,117.82,1019.29,1526.21,2.96,10.60
20,389.40,116.37,1025.98,1531.75,2.97,10.64
21,389.23,117.68,1006.54,1513.45,2.93,10.51
22,386.03,118.10,1002.90,1507.04,2.92,10.47
23,390.48,118.45,1008.70,1517.63,2.94,10.54
24,383.99,117.77,1020.78,1522.55,2.95,10.57
25,384.95,118.49,1015.70,1519.14,2.94,10.55
26,385.23,118.98,1010.46,1514.67,2.93,10.52
27,388.10,116.80,1000.01,1504.91,2.92,10.45
28,383.48,118.86,1019.36,1521.69,2.95,10.57
29,387.95,116.42,1008.53,1512.89,2.93,10.51
30,383.55,118.44,1012.36,1514.34,2.93,10.52
31,386.61,117.27,1016.40,1520.28,2.95,10.56
profile_csv=sevenb-ffn-kernel-ab/baseline-1.csv
