[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (317092 > 262144). Running this sequence through the model will result in indexing errors
[i8] basket: 10 prompts x 16384 tok, domains ['code', 'longctx', 'math', 'sci', 'wiki']
boot: this card's residency could NOT be split between this process and any other — neither NVML's per-process walk nor a pre-context mark was available — so driver.cuda_context is being booked from the DEVICE-WIDE reading (397 MiB). If anything else is resident on this device, its bytes are inside that number and the KV pool is sized that much smaller for the life of this boot.
compile-cache MISS: xgrammar — xgrammar compiled grammars (re)compiles this boot [/cache/xgrammar]
VRAM at profile: this card's residency could NOT be split between this process and any other — NVML's per-process walk did not identify us — so the whole 14.26 GiB resident on device 0 is being booked as driver.cuda_context. Any co-tenant's bytes are inside that number and KV is sized that much smaller for this boot.
[TKV autotune-provenance] fp=57911e26f4e47f74 regime=cache-hit cells=8 sweep_src=2772549a0ef713db tile_tokens=4 num_splits=32,64,128,256 min_blocks_per_sm=0,3 resolved=0 held=0 inherited=0
activation profile: the decode probe's FIRST call cost 13301 allocator events / 1 MiB / 138.1 ms against 13305 / 1 MiB / 52.9 ms warm — a first-call kernel search inside the probe, not a step. The reported peak AND device time are the warm ones (3 passes, the last is what is reported); the cold numbers are logged so the difference is visible.
driver.modules_loaded baseline: SEEDED at 144703488 B (0.135 GiB) → /root/.cache/arbi-serve/budget-cache/69b5561ff6f9c4cf.modules.json. This boot ran UNGATED — it booked its own bracketed growth, so an unregistered pool or a raw cudaMalloc would be inside that number rather than on driver.residual. Every later boot at this configuration is held to it. Expected ONCE per configuration; if it repeats, the budget cache is not persisting (point ARBI_SERVE_BUDGET_CACHE_DIR at a durable volume) and the guard is inert.
/opt/venv/lib/python3.12/site-packages/torch/__init__.py:1172: FutureWarning: `torch.distributed.reduce_op` is deprecated, please use `torch.distributed.ReduceOp` instead
  return isinstance(obj, torch.Tensor)
[i8] trellis->path map: 401 linears
JIT compile AFTER serving-ready [cpp_ext]: exl3_i8_gemm cached .so load — a live request paid this compile's latency. This is a boot-warmup coverage gap: extend warmup to pre-compile this kernel/specialization. Counter jit_compile_serving (must-not-fire) at GET /v1/admin/flag_truth.
[i8] prefix cache: server default resolved=True, per-request cache_enabled=False; any cache hit below is refused
driver serving-growth OVER RESERVE: the driver holds 2 MiB more than it did when the boot brackets closed, against the 0 MiB the serving floor's transient.serving_step.driver_growth term held for it. Those bytes are in no pool and no allocator counter -- the live card reaches them as driver.modules_serving and the boot ledger only as driver.residual -- and the KV layout is frozen, so the overage comes out of the free VRAM the floor is holding for one in-flight step (the verify tail, the DFlash context-assemble) and those reserves are the ones that fail first. The reading is on record for this configuration, so the next boot holds it; this process serves the rest of its life short by the overage.
activation admission RE-ARMED (driver residency over reserve): the step budget FELL to 5751 MiB, -642 MiB on the 6393 MiB it was enforcing and 6393 MiB the boot reading armed. The widest slate admission will build is narrower from here — narrower prefill chunks and deferred rows instead of a forward that OOMs after the layout freeze.
[i8] warmup done
JIT compile AFTER serving-ready [cpp_ext]: tkv_unified_splitk_g6_v8_k4_v4_m1_v1_smem_rngfull_lbmb3_gms1_mw8_tt4_sh1_rc4_pf1_ldbypass cached .so load — a live request paid this compile's latency. This is a boot-warmup coverage gap: extend warmup to pre-compile this kernel/specialization. Counter jit_compile_serving (must-not-fire) at GET /v1/admin/flag_truth.
  [0] code           REF      16384tok  wall=  5.05s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [0] code           REF2     16384tok  wall=  5.01s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [0] code           STD      16384tok  wall=  5.29s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [0] code           REF32    16384tok  wall=  6.80s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [0] code           STD32    16384tok  wall=  7.06s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [0] code           INT8     16384tok  wall=  3.78s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [1] code           REF      16384tok  wall=  5.06s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [1] code           REF2     16384tok  wall=  5.07s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [1] code           STD      16384tok  wall=  5.32s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [1] code           REF32    16384tok  wall=  6.83s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [1] code           STD32    16384tok  wall=  7.07s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [1] code           INT8     16384tok  wall=  3.78s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [2] wiki           REF      16384tok  wall=  5.08s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [2] wiki           REF2     16384tok  wall=  5.08s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [2] wiki           STD      16384tok  wall=  5.34s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [2] wiki           REF32    16384tok  wall=  6.85s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [2] wiki           STD32    16384tok  wall=  7.09s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [2] wiki           INT8     16384tok  wall=  3.79s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [3] wiki           REF      16384tok  wall=  5.09s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [3] wiki           REF2     16384tok  wall=  5.09s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [3] wiki           STD      16384tok  wall=  5.34s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [3] wiki           REF32    16384tok  wall=  6.85s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [3] wiki           STD32    16384tok  wall=  7.10s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [3] wiki           INT8     16384tok  wall=  3.80s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [4] math           REF      16384tok  wall=  5.10s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [4] math           REF2     16384tok  wall=  5.11s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [4] math           STD      16384tok  wall=  5.35s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [4] math           REF32    16384tok  wall=  6.86s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [4] math           STD32    16384tok  wall=  7.10s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [4] math           INT8     16384tok  wall=  3.80s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [5] math           REF      16384tok  wall=  5.12s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [5] math           REF2     16384tok  wall=  5.10s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [5] math           STD      16384tok  wall=  5.38s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [5] math           REF32    16384tok  wall=  6.89s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [5] math           STD32    16384tok  wall=  7.11s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [5] math           INT8     16384tok  wall=  3.80s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [6] longctx        REF      16384tok  wall=  5.14s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [6] longctx        REF2     16384tok  wall=  5.13s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [6] longctx        STD      16384tok  wall=  5.38s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [6] longctx        REF32    16384tok  wall=  6.89s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [6] longctx        STD32    16384tok  wall=  7.13s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [6] longctx        INT8     16384tok  wall=  3.81s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [7] longctx        REF      16384tok  wall=  5.13s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [7] longctx        REF2     16384tok  wall=  5.13s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [7] longctx        STD      16384tok  wall=  5.38s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [7] longctx        REF32    16384tok  wall=  6.88s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [7] longctx        STD32    16384tok  wall=  7.12s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [7] longctx        INT8     16384tok  wall=  3.81s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [8] sci            REF      16384tok  wall=  5.13s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [8] sci            REF2     16384tok  wall=  5.14s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [8] sci            STD      16384tok  wall=  5.38s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [8] sci            REF32    16384tok  wall=  6.87s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [8] sci            STD32    16384tok  wall=  7.11s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [8] sci            INT8     16384tok  wall=  3.81s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [9] sci            REF      16384tok  wall=  5.11s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [9] sci            REF2     16384tok  wall=  5.11s  finish=length  leg=3200F/0S acc=3200x16/0x32
  [9] sci            STD      16384tok  wall=  5.37s  finish=length  leg=0F/3200S acc=3200x16/0x32
  [9] sci            REF32    16384tok  wall=  6.86s  finish=length  leg=3200F/0S acc=0x16/3200x32
  [9] sci            STD32    16384tok  wall=  7.11s  finish=length  leg=0F/3200S acc=0x16/3200x32
  [9] sci            INT8     16384tok  wall=  3.81s  finish=length  leg=0F/3200S acc=3200x16/0x32

[i8] kernel census: int8 calls=32000 rows=65536000 reconstructs SKIPPED=32000 fallback (non-4bpw) calls=0
[i8] shapes served by the kernel (K,N)->calls: {(5120, 1024): 2560, (5120, 6144): 3840, (5120, 10240): 3840, (5120, 12288): 1280, (5120, 17408): 10240, (6144, 5120): 5120, (17408, 5120): 5120}
[i8] excluded projections: none  excluded calls=0

  linear class                                           served   excl  int8 vs fp32 legB vs fp32 int8 vs legB
  model.layers.0.linear_attn.in_proj_qkv                     81      0     4.827e-03    1.713e-03    5.132e-03
  model.layers.0.linear_attn.in_proj_z                       81      0     6.316e-03    1.129e-03    6.418e-03
  model.layers.0.linear_attn.out_proj                        81      0     1.249e-02    2.189e-03    1.268e-02
  model.layers.0.mlp.down_proj                               81      0     2.122e-02    2.609e-03    2.137e-02
  model.layers.0.mlp.gate_proj                               81      0     9.121e-03    3.413e-03    9.739e-03
  model.layers.0.mlp.up_proj                                 81      0     9.082e-03    3.418e-03    9.702e-03
  model.layers.1.linear_attn.in_proj_qkv                     81      0     1.186e-02    2.166e-03    1.206e-02
  model.layers.1.linear_attn.in_proj_z                       81      0     1.361e-02    1.840e-03    1.373e-02
  model.layers.1.linear_attn.out_proj                        81      0     1.604e-02    2.254e-03    1.619e-02
  model.layers.1.mlp.down_proj                               81      0     1.344e-02    2.322e-03    1.364e-02
  model.layers.1.mlp.gate_proj                               81      0     9.700e-03    3.460e-03    1.030e-02
  model.layers.1.mlp.up_proj                                 81      0     1.015e-02    3.488e-03    1.073e-02
  model.layers.10.linear_attn.in_proj_qkv                    81      0     1.304e-02    2.255e-03    1.324e-02
  model.layers.10.linear_attn.in_proj_z                      81      0     1.298e-02    1.836e-03    1.311e-02
  model.layers.10.linear_attn.out_proj                       81      0     1.497e-02    2.120e-03    1.512e-02
  model.layers.10.mlp.down_proj                              81      0     1.268e-02    2.348e-03    1.289e-02
  model.layers.10.mlp.gate_proj                              81      0     1.035e-02    3.581e-03    1.095e-02
  model.layers.10.mlp.up_proj                                81      0     1.076e-02    3.588e-03    1.134e-02
  model.layers.11.mlp.down_proj                              81      0     1.333e-02    2.506e-03    1.356e-02
  model.layers.11.mlp.gate_proj                              81      0     1.027e-02    3.507e-03    1.085e-02
  model.layers.11.mlp.up_proj                                81      0     1.088e-02    3.576e-03    1.145e-02
  model.layers.11.self_attn.k_proj                           81      0     1.330e-02    2.705e-03    1.357e-02
  model.layers.11.self_attn.o_proj                           81      0     1.304e-02    2.370e-03    1.326e-02
  model.layers.11.self_attn.q_proj                           81      0     1.174e-02    2.022e-03    1.191e-02
  model.layers.11.self_attn.v_proj                           81      0     1.037e-02    2.283e-03    1.062e-02
  model.layers.12.linear_attn.in_proj_qkv                    81      0     1.292e-02    2.244e-03    1.312e-02
  model.layers.12.linear_attn.in_proj_z                      81      0     1.365e-02    1.892e-03    1.378e-02
  model.layers.12.linear_attn.out_proj                       81      0     1.583e-02    2.138e-03    1.597e-02
  model.layers.12.mlp.down_proj                              81      0     1.316e-02    2.550e-03    1.341e-02
  model.layers.12.mlp.gate_proj                              81      0     1.002e-02    3.472e-03    1.061e-02
  model.layers.12.mlp.up_proj                                81      0     1.072e-02    3.553e-03    1.129e-02
  model.layers.13.linear_attn.in_proj_qkv                    81      0     1.292e-02    2.241e-03    1.312e-02
  model.layers.13.linear_attn.in_proj_z                      81      0     1.418e-02    1.970e-03    1.431e-02
  model.layers.13.linear_attn.out_proj                       81      0     1.680e-02    2.175e-03    1.693e-02
  model.layers.13.mlp.down_proj                              81      0     1.373e-02    2.565e-03    1.396e-02
  model.layers.13.mlp.gate_proj                              81      0     1.007e-02    3.483e-03    1.066e-02
  model.layers.13.mlp.up_proj                                81      0     1.071e-02    3.557e-03    1.128e-02
  model.layers.14.linear_attn.in_proj_qkv                    81      0     1.272e-02    2.232e-03    1.291e-02
  model.layers.14.linear_attn.in_proj_z                      81      0     1.334e-02    1.883e-03    1.348e-02
  model.layers.14.linear_attn.out_proj                       81      0     1.586e-02    2.214e-03    1.601e-02
  model.layers.14.mlp.down_proj                              81      0     1.300e-02    2.463e-03    1.323e-02
  model.layers.14.mlp.gate_proj                              81      0     9.963e-03    3.503e-03    1.056e-02
  model.layers.14.mlp.up_proj                                81      0     1.064e-02    3.565e-03    1.122e-02
  model.layers.15.mlp.down_proj                              81      0     1.349e-02    2.537e-03    1.373e-02
  model.layers.15.mlp.gate_proj                              81      0     1.010e-02    3.488e-03    1.068e-02
  model.layers.15.mlp.up_proj                                81      0     1.068e-02    3.562e-03    1.125e-02
  model.layers.15.self_attn.k_proj                           81      0     1.452e-02    2.890e-03    1.481e-02
  model.layers.15.self_attn.o_proj                           81      0     1.398e-02    2.117e-03    1.414e-02
  model.layers.15.self_attn.q_proj                           81      0     1.302e-02    2.087e-03    1.318e-02
  model.layers.15.self_attn.v_proj                           81      0     9.832e-03    2.228e-03    1.008e-02
  model.layers.16.linear_attn.in_proj_qkv                    81      0     1.289e-02    2.246e-03    1.308e-02
  model.layers.16.linear_attn.in_proj_z                      81      0     1.353e-02    1.895e-03    1.366e-02
  model.layers.16.linear_attn.out_proj                       81      0     1.636e-02    2.275e-03    1.651e-02
  model.layers.16.mlp.down_proj                              81      0     1.298e-02    2.519e-03    1.322e-02
  model.layers.16.mlp.gate_proj                              81      0     1.006e-02    3.472e-03    1.064e-02
  model.layers.16.mlp.up_proj                                81      0     1.070e-02    3.563e-03    1.128e-02
  model.layers.17.linear_attn.in_proj_qkv                    81      0     1.287e-02    2.222e-03    1.306e-02
  model.layers.17.linear_attn.in_proj_z                      81      0     1.352e-02    1.896e-03    1.366e-02
  model.layers.17.linear_attn.out_proj                       81      0     1.781e-02    2.245e-03    1.796e-02
  model.layers.17.mlp.down_proj                              81      0     1.381e-02    2.455e-03    1.403e-02
  model.layers.17.mlp.gate_proj                              81      0     1.010e-02    3.471e-03    1.068e-02
  model.layers.17.mlp.up_proj                                81      0     1.072e-02    3.565e-03    1.129e-02
  model.layers.18.linear_attn.in_proj_qkv                    81      0     1.206e-02    2.189e-03    1.225e-02
  model.layers.18.linear_attn.in_proj_z                      81      0     1.264e-02    1.788e-03    1.276e-02
  model.layers.18.linear_attn.out_proj                       81      0     1.527e-02    2.227e-03    1.543e-02
  model.layers.18.mlp.down_proj                              81      0     1.461e-02    2.424e-03    1.481e-02
  model.layers.18.mlp.gate_proj                              81      0     1.024e-02    3.422e-03    1.079e-02
  model.layers.18.mlp.up_proj                                81      0     1.058e-02    3.532e-03    1.116e-02
  model.layers.19.mlp.down_proj                              81      0     1.585e-02    2.536e-03    1.605e-02
  model.layers.19.mlp.gate_proj                              81      0     1.008e-02    3.395e-03    1.063e-02
  model.layers.19.mlp.up_proj                                81      0     1.096e-02    3.581e-03    1.153e-02
  model.layers.19.self_attn.k_proj                           81      0     1.407e-02    2.788e-03    1.434e-02
  model.layers.19.self_attn.o_proj                           81      0     1.814e-02    2.166e-03    1.826e-02
  model.layers.19.self_attn.q_proj                           81      0     9.500e-03    1.927e-03    9.692e-03
  model.layers.19.self_attn.v_proj                           81      0     1.047e-02    2.352e-03    1.073e-02
  model.layers.2.linear_attn.in_proj_qkv                     81      0     1.249e-02    2.213e-03    1.268e-02
  model.layers.2.linear_attn.in_proj_z                       81      0     1.398e-02    1.926e-03    1.411e-02
  model.layers.2.linear_attn.out_proj                        81      0     1.794e-02    2.322e-03    1.808e-02
  model.layers.2.mlp.down_proj                               81      0     1.706e-02    2.372e-03    1.723e-02
  model.layers.2.mlp.gate_proj                               81      0     1.029e-02    3.530e-03    1.088e-02
  model.layers.2.mlp.up_proj                                 81      0     1.061e-02    3.561e-03    1.120e-02
  model.layers.20.linear_attn.in_proj_qkv                    81      0     1.215e-02    2.138e-03    1.234e-02
  model.layers.20.linear_attn.in_proj_z                      81      0     1.272e-02    1.791e-03    1.285e-02
  model.layers.20.linear_attn.out_proj                       81      0     1.578e-02    2.229e-03    1.594e-02
  model.layers.20.mlp.down_proj                              81      0     1.592e-02    2.582e-03    1.613e-02
  model.layers.20.mlp.gate_proj                              81      0     1.009e-02    3.407e-03    1.065e-02
  model.layers.20.mlp.up_proj                                81      0     1.128e-02    3.604e-03    1.184e-02
  model.layers.21.linear_attn.in_proj_qkv                    81      0     1.198e-02    2.201e-03    1.218e-02
  model.layers.21.linear_attn.in_proj_z                      81      0     1.162e-02    1.671e-03    1.174e-02
  model.layers.21.linear_attn.out_proj                       81      0     1.703e-02    2.236e-03    1.717e-02
  model.layers.21.mlp.down_proj                              81      0     1.719e-02    2.592e-03    1.739e-02
  model.layers.21.mlp.gate_proj                              81      0     1.010e-02    3.421e-03    1.066e-02
  model.layers.21.mlp.up_proj                                81      0     1.124e-02    3.609e-03    1.180e-02
  model.layers.22.linear_attn.in_proj_qkv                    81      0     1.201e-02    2.240e-03    1.221e-02
  model.layers.22.linear_attn.in_proj_z                      81      0     1.202e-02    1.714e-03    1.214e-02
  model.layers.22.linear_attn.out_proj                       81      0     1.580e-02    2.304e-03    1.597e-02
  model.layers.22.mlp.down_proj                              81      0     1.511e-02    2.552e-03    1.533e-02
  model.layers.22.mlp.gate_proj                              81      0     1.054e-02    3.140e-03    1.100e-02
  model.layers.22.mlp.up_proj                                81      0     1.140e-02    3.621e-03    1.196e-02
  model.layers.23.mlp.down_proj                              81      0     1.568e-02    2.553e-03    1.589e-02
  model.layers.23.mlp.gate_proj                              81      0     1.009e-02    3.367e-03    1.063e-02
  model.layers.23.mlp.up_proj                                81      0     1.141e-02    3.621e-03    1.197e-02
  model.layers.23.self_attn.k_proj                           81      0     1.382e-02    2.675e-03    1.408e-02
  model.layers.23.self_attn.o_proj                           81      0     1.681e-02    2.423e-03    1.698e-02
  model.layers.23.self_attn.q_proj                           81      0     9.476e-03    1.964e-03    9.678e-03
  model.layers.23.self_attn.v_proj                           81      0     1.034e-02    2.335e-03    1.061e-02
  model.layers.24.linear_attn.in_proj_qkv                    81      0     1.188e-02    2.253e-03    1.210e-02
  model.layers.24.linear_attn.in_proj_z                      81      0     1.223e-02    1.734e-03    1.235e-02
  model.layers.24.linear_attn.out_proj                       81      0     1.604e-02    2.270e-03    1.620e-02
  model.layers.24.mlp.down_proj                              81      0     1.459e-02    2.620e-03    1.482e-02
  model.layers.24.mlp.gate_proj                              81      0     1.026e-02    3.316e-03    1.078e-02
  model.layers.24.mlp.up_proj                                81      0     1.144e-02    3.612e-03    1.199e-02
  model.layers.25.linear_attn.in_proj_qkv                    81      0     1.141e-02    2.212e-03    1.162e-02
  model.layers.25.linear_attn.in_proj_z                      81      0     1.279e-02    1.788e-03    1.291e-02
  model.layers.25.linear_attn.out_proj                       81      0     1.475e-02    2.178e-03    1.491e-02
  model.layers.25.mlp.down_proj                              81      0     1.391e-02    2.553e-03    1.414e-02
  model.layers.25.mlp.gate_proj                              81      0     1.054e-02    3.323e-03    1.105e-02
  model.layers.25.mlp.up_proj                                81      0     1.130e-02    3.601e-03    1.185e-02
  model.layers.26.linear_attn.in_proj_qkv                    81      0     1.151e-02    2.236e-03    1.172e-02
  model.layers.26.linear_attn.in_proj_z                      81      0     1.259e-02    1.796e-03    1.272e-02
  model.layers.26.linear_attn.out_proj                       81      0     1.494e-02    2.342e-03    1.512e-02
  model.layers.26.mlp.down_proj                              81      0     1.268e-02    2.467e-03    1.292e-02
  model.layers.26.mlp.gate_proj                              81      0     1.112e-02    3.295e-03    1.160e-02
  model.layers.26.mlp.up_proj                                81      0     1.085e-02    3.549e-03    1.142e-02
  model.layers.27.mlp.down_proj                              81      0     1.335e-02    2.577e-03    1.360e-02
  model.layers.27.mlp.gate_proj                              81      0     1.080e-02    3.204e-03    1.126e-02
  model.layers.27.mlp.up_proj                                81      0     1.090e-02    3.553e-03    1.147e-02
  model.layers.27.self_attn.k_proj                           81      0     1.215e-02    2.501e-03    1.240e-02
  model.layers.27.self_attn.o_proj                           81      0     1.711e-02    2.742e-03    1.733e-02
  model.layers.27.self_attn.q_proj                           81      0     1.021e-02    2.059e-03    1.042e-02
  model.layers.27.self_attn.v_proj                           81      0     9.828e-03    2.250e-03    1.009e-02
  model.layers.28.linear_attn.in_proj_qkv                    81      0     1.179e-02    2.219e-03    1.200e-02
  model.layers.28.linear_attn.in_proj_z                      81      0     1.290e-02    1.818e-03    1.302e-02
  model.layers.28.linear_attn.out_proj                       81      0     1.477e-02    2.298e-03    1.495e-02
  model.layers.28.mlp.down_proj                              81      0     1.301e-02    2.596e-03    1.327e-02
  model.layers.28.mlp.gate_proj                              81      0     1.104e-02    3.175e-03    1.149e-02
  model.layers.28.mlp.up_proj                                81      0     1.084e-02    3.538e-03    1.141e-02
  model.layers.29.linear_attn.in_proj_qkv                    81      0     1.205e-02    2.246e-03    1.226e-02
  model.layers.29.linear_attn.in_proj_z                      81      0     1.395e-02    1.964e-03    1.408e-02
  model.layers.29.linear_attn.out_proj                       81      0     1.658e-02    2.351e-03    1.674e-02
  model.layers.29.mlp.down_proj                              81      0     1.346e-02    2.623e-03    1.371e-02
  model.layers.29.mlp.gate_proj                              81      0     1.115e-02    3.186e-03    1.159e-02
  model.layers.29.mlp.up_proj                                81      0     1.080e-02    3.540e-03    1.137e-02
  model.layers.3.mlp.down_proj                               81      0     1.415e-02    2.578e-03    1.438e-02
  model.layers.3.mlp.gate_proj                               81      0     1.017e-02    3.527e-03    1.076e-02
  model.layers.3.mlp.up_proj                                 81      0     1.100e-02    3.601e-03    1.158e-02
  model.layers.3.self_attn.k_proj                            81      0     1.247e-02    2.581e-03    1.273e-02
  model.layers.3.self_attn.o_proj                            81      0     1.495e-02    2.177e-03    1.511e-02
  model.layers.3.self_attn.q_proj                            81      0     1.169e-02    2.033e-03    1.187e-02
  model.layers.3.self_attn.v_proj                            81      0     1.263e-02    2.305e-03    1.285e-02
  model.layers.30.linear_attn.in_proj_qkv                    81      0     1.190e-02    2.231e-03    1.211e-02
  model.layers.30.linear_attn.in_proj_z                      81      0     1.349e-02    1.888e-03    1.362e-02
  model.layers.30.linear_attn.out_proj                       81      0     1.621e-02    2.413e-03    1.639e-02
  model.layers.30.mlp.down_proj                              81      0     1.287e-02    2.561e-03    1.312e-02
  model.layers.30.mlp.gate_proj                              81      0     1.172e-02    3.189e-03    1.214e-02
  model.layers.30.mlp.up_proj                                81      0     1.067e-02    3.520e-03    1.123e-02
  model.layers.31.mlp.down_proj                              81      0     1.317e-02    2.594e-03    1.342e-02
  model.layers.31.mlp.gate_proj                              81      0     1.148e-02    3.154e-03    1.191e-02
  model.layers.31.mlp.up_proj                                81      0     1.090e-02    3.518e-03    1.145e-02
  model.layers.31.self_attn.k_proj                           81      0     1.207e-02    2.525e-03    1.233e-02
  model.layers.31.self_attn.o_proj                           81      0     2.213e-02    3.026e-03    2.233e-02
  model.layers.31.self_attn.q_proj                           81      0     1.016e-02    2.073e-03    1.037e-02
  model.layers.31.self_attn.v_proj                           81      0     9.914e-03    2.186e-03    1.014e-02
  model.layers.32.linear_attn.in_proj_qkv                    81      0     1.282e-02    2.288e-03    1.302e-02
  model.layers.32.linear_attn.in_proj_z                      81      0     1.412e-02    1.983e-03    1.425e-02
  model.layers.32.linear_attn.out_proj                       81      0     1.564e-02    2.301e-03    1.581e-02
  model.layers.32.mlp.down_proj                              81      0     1.303e-02    2.618e-03    1.329e-02
  model.layers.32.mlp.gate_proj                              81      0     1.166e-02    3.107e-03    1.206e-02
  model.layers.32.mlp.up_proj                                81      0     1.108e-02    3.524e-03    1.163e-02
  model.layers.33.linear_attn.in_proj_qkv                    81      0     1.261e-02    2.298e-03    1.282e-02
  model.layers.33.linear_attn.in_proj_z                      81      0     1.385e-02    1.990e-03    1.399e-02
  model.layers.33.linear_attn.out_proj                       81      0     1.747e-02    2.364e-03    1.763e-02
  model.layers.33.mlp.down_proj                              81      0     1.343e-02    2.602e-03    1.368e-02
  model.layers.33.mlp.gate_proj                              81      0     1.191e-02    3.112e-03    1.231e-02
  model.layers.33.mlp.up_proj                                81      0     1.124e-02    3.540e-03    1.178e-02
  model.layers.34.linear_attn.in_proj_qkv                    81      0     1.252e-02    2.276e-03    1.273e-02
  model.layers.34.linear_attn.in_proj_z                      81      0     1.341e-02    1.911e-03    1.355e-02
  model.layers.34.linear_attn.out_proj                       81      0     1.617e-02    2.364e-03    1.634e-02
  model.layers.34.mlp.down_proj                              81      0     1.379e-02    2.611e-03    1.404e-02
  model.layers.34.mlp.gate_proj                              81      0     1.206e-02    3.070e-03    1.244e-02
  model.layers.34.mlp.up_proj                                81      0     1.115e-02    3.506e-03    1.169e-02
  model.layers.35.mlp.down_proj                              81      0     1.417e-02    2.560e-03    1.441e-02
  model.layers.35.mlp.gate_proj                              81      0     1.066e-02    3.067e-03    1.109e-02
  model.layers.35.mlp.up_proj                                81      0     1.138e-02    3.579e-03    1.193e-02
  model.layers.35.self_attn.k_proj                           81      0     1.175e-02    2.504e-03    1.202e-02
  model.layers.35.self_attn.o_proj                           81      0     1.991e-02    2.804e-03    2.011e-02
  model.layers.35.self_attn.q_proj                           81      0     1.072e-02    2.115e-03    1.093e-02
  model.layers.35.self_attn.v_proj                           81      0     1.073e-02    2.262e-03    1.096e-02
  model.layers.36.linear_attn.in_proj_qkv                    81      0     1.294e-02    2.274e-03    1.314e-02
  model.layers.36.linear_attn.in_proj_z                      81      0     1.210e-02    1.745e-03    1.223e-02
  model.layers.36.linear_attn.out_proj                       81      0     1.662e-02    2.261e-03    1.678e-02
  model.layers.36.mlp.down_proj                              81      0     1.508e-02    2.627e-03    1.531e-02
  model.layers.36.mlp.gate_proj                              81      0     1.025e-02    3.143e-03    1.072e-02
  model.layers.36.mlp.up_proj                                81      0     1.154e-02    3.605e-03    1.209e-02
  model.layers.37.linear_attn.in_proj_qkv                    81      0     1.223e-02    2.241e-03    1.243e-02
  model.layers.37.linear_attn.in_proj_z                      81      0     1.059e-02    1.597e-03    1.071e-02
  model.layers.37.linear_attn.out_proj                       81      0     1.839e-02    2.197e-03    1.852e-02
  model.layers.37.mlp.down_proj                              81      0     1.540e-02    2.731e-03    1.564e-02
  model.layers.37.mlp.gate_proj                              81      0     1.056e-02    3.237e-03    1.104e-02
  model.layers.37.mlp.up_proj                                81      0     1.173e-02    3.628e-03    1.228e-02
  model.layers.38.linear_attn.in_proj_qkv                    81      0     1.200e-02    2.317e-03    1.222e-02
  model.layers.38.linear_attn.in_proj_z                      81      0     1.099e-02    1.628e-03    1.111e-02
  model.layers.38.linear_attn.out_proj                       81      0     1.656e-02    2.310e-03    1.672e-02
  model.layers.38.mlp.down_proj                              81      0     1.595e-02    2.699e-03    1.617e-02
  model.layers.38.mlp.gate_proj                              81      0     1.069e-02    3.068e-03    1.112e-02
  model.layers.38.mlp.up_proj                                81      0     1.171e-02    3.638e-03    1.226e-02
  model.layers.39.mlp.down_proj                              81      0     1.439e-02    2.656e-03    1.463e-02
  model.layers.39.mlp.gate_proj                              81      0     9.970e-03    3.161e-03    1.046e-02
  model.layers.39.mlp.up_proj                                81      0     1.179e-02    3.647e-03    1.234e-02
  model.layers.39.self_attn.k_proj                           81      0     1.178e-02    2.463e-03    1.204e-02
  model.layers.39.self_attn.o_proj                           81      0     2.078e-02    2.837e-03    2.097e-02
  model.layers.39.self_attn.q_proj                           81      0     1.046e-02    2.097e-03    1.067e-02
  model.layers.39.self_attn.v_proj                           81      0     1.091e-02    2.328e-03    1.116e-02
  model.layers.4.linear_attn.in_proj_qkv                     81      0     1.192e-02    2.308e-03    1.214e-02
  model.layers.4.linear_attn.in_proj_z                       81      0     1.247e-02    1.771e-03    1.259e-02
  model.layers.4.linear_attn.out_proj                        81      0     1.789e-02    2.411e-03    1.806e-02
  model.layers.4.mlp.down_proj                               81      0     1.443e-02    2.603e-03    1.467e-02
  model.layers.4.mlp.gate_proj                               81      0     1.013e-02    3.511e-03    1.072e-02
  model.layers.4.mlp.up_proj                                 81      0     1.123e-02    3.618e-03    1.180e-02
  model.layers.40.linear_attn.in_proj_qkv                    81      0     1.224e-02    2.282e-03    1.245e-02
  model.layers.40.linear_attn.in_proj_z                      81      0     1.173e-02    1.688e-03    1.185e-02
  model.layers.40.linear_attn.out_proj                       81      0     1.607e-02    2.268e-03    1.623e-02
  model.layers.40.mlp.down_proj                              81      0     1.478e-02    2.675e-03    1.501e-02
  model.layers.40.mlp.gate_proj                              81      0     1.008e-02    3.134e-03    1.055e-02
  model.layers.40.mlp.up_proj                                81      0     1.164e-02    3.641e-03    1.220e-02
  model.layers.41.linear_attn.in_proj_qkv                    81      0     1.176e-02    2.256e-03    1.197e-02
  model.layers.41.linear_attn.in_proj_z                      81      0     1.226e-02    1.733e-03    1.238e-02
  model.layers.41.linear_attn.out_proj                       81      0     1.649e-02    2.255e-03    1.664e-02
  model.layers.41.mlp.down_proj                              81      0     1.477e-02    2.705e-03    1.502e-02
  model.layers.41.mlp.gate_proj                              81      0     1.085e-02    3.211e-03    1.132e-02
  model.layers.41.mlp.up_proj                                81      0     1.167e-02    3.636e-03    1.223e-02
  model.layers.42.linear_attn.in_proj_qkv                    81      0     1.162e-02    2.276e-03    1.184e-02
  model.layers.42.linear_attn.in_proj_z                      81      0     1.242e-02    1.747e-03    1.254e-02
  model.layers.42.linear_attn.out_proj                       81      0     1.568e-02    2.272e-03    1.584e-02
  model.layers.42.mlp.down_proj                              81      0     1.467e-02    2.652e-03    1.490e-02
  model.layers.42.mlp.gate_proj                              81      0     1.173e-02    3.240e-03    1.217e-02
  model.layers.42.mlp.up_proj                                81      0     1.143e-02    3.586e-03    1.198e-02
  model.layers.43.mlp.down_proj                              81      0     1.386e-02    2.670e-03    1.411e-02
  model.layers.43.mlp.gate_proj                              81      0     1.084e-02    3.218e-03    1.131e-02
  model.layers.43.mlp.up_proj                                81      0     1.153e-02    3.615e-03    1.208e-02
  model.layers.43.self_attn.k_proj                           81      0     1.250e-02    2.453e-03    1.274e-02
  model.layers.43.self_attn.o_proj                           81      0     2.162e-02    3.131e-03    2.185e-02
  model.layers.43.self_attn.q_proj                           81      0     1.059e-02    2.086e-03    1.079e-02
  model.layers.43.self_attn.v_proj                           81      0     1.050e-02    2.259e-03    1.074e-02
  model.layers.44.linear_attn.in_proj_qkv                    81      0     1.203e-02    2.229e-03    1.224e-02
  model.layers.44.linear_attn.in_proj_z                      81      0     1.227e-02    1.697e-03    1.239e-02
  model.layers.44.linear_attn.out_proj                       81      0     1.629e-02    2.299e-03    1.645e-02
  model.layers.44.mlp.down_proj                              81      0     1.413e-02    2.713e-03    1.439e-02
  model.layers.44.mlp.gate_proj                              81      0     1.082e-02    3.260e-03    1.130e-02
  model.layers.44.mlp.up_proj                                81      0     1.171e-02    3.618e-03    1.225e-02
  model.layers.45.linear_attn.in_proj_qkv                    81      0     1.207e-02    2.277e-03    1.229e-02
  model.layers.45.linear_attn.in_proj_z                      81      0     1.264e-02    1.761e-03    1.276e-02
  model.layers.45.linear_attn.out_proj                       81      0     1.809e-02    2.431e-03    1.825e-02
  model.layers.45.mlp.down_proj                              81      0     1.435e-02    2.745e-03    1.461e-02
  model.layers.45.mlp.gate_proj                              81      0     1.081e-02    3.287e-03    1.130e-02
  model.layers.45.mlp.up_proj                                81      0     1.157e-02    3.622e-03    1.212e-02
  model.layers.46.linear_attn.in_proj_qkv                    81      0     1.197e-02    2.246e-03    1.218e-02
  model.layers.46.linear_attn.in_proj_z                      81      0     1.186e-02    1.650e-03    1.197e-02
  model.layers.46.linear_attn.out_proj                       81      0     1.870e-02    2.364e-03    1.885e-02
  model.layers.46.mlp.down_proj                              81      0     1.426e-02    2.686e-03    1.451e-02
  model.layers.46.mlp.gate_proj                              81      0     1.112e-02    3.310e-03    1.160e-02
  model.layers.46.mlp.up_proj                                81      0     1.150e-02    3.598e-03    1.205e-02
  model.layers.47.mlp.down_proj                              81      0     1.427e-02    2.639e-03    1.451e-02
  model.layers.47.mlp.gate_proj                              81      0     1.056e-02    3.336e-03    1.107e-02
  model.layers.47.mlp.up_proj                                81      0     1.177e-02    3.624e-03    1.232e-02
  model.layers.47.self_attn.k_proj                           81      0     1.243e-02    2.468e-03    1.268e-02
  model.layers.47.self_attn.o_proj                           81      0     1.958e-02    2.554e-03    1.975e-02
  model.layers.47.self_attn.q_proj                           81      0     9.645e-03    2.054e-03    9.860e-03
  model.layers.47.self_attn.v_proj                           81      0     1.061e-02    2.234e-03    1.084e-02
  model.layers.48.linear_attn.in_proj_qkv                    81      0     1.272e-02    2.266e-03    1.292e-02
  model.layers.48.linear_attn.in_proj_z                      81      0     1.209e-02    1.649e-03    1.220e-02
  model.layers.48.linear_attn.out_proj                       81      0     1.710e-02    2.416e-03    1.727e-02
  model.layers.48.mlp.down_proj                              81      0     1.481e-02    2.681e-03    1.505e-02
  model.layers.48.mlp.gate_proj                              81      0     1.035e-02    3.332e-03    1.087e-02
  model.layers.48.mlp.up_proj                                81      0     1.169e-02    3.605e-03    1.223e-02
  model.layers.49.linear_attn.in_proj_qkv                    81      0     1.209e-02    2.281e-03    1.230e-02
  model.layers.49.linear_attn.in_proj_z                      81      0     1.152e-02    1.601e-03    1.163e-02
  model.layers.49.linear_attn.out_proj                       81      0     1.905e-02    2.170e-03    1.918e-02
  model.layers.49.mlp.down_proj                              81      0     1.568e-02    2.694e-03    1.591e-02
  model.layers.49.mlp.gate_proj                              81      0     1.006e-02    3.369e-03    1.061e-02
  model.layers.49.mlp.up_proj                                81      0     1.149e-02    3.564e-03    1.203e-02
  model.layers.5.linear_attn.in_proj_qkv                     81      0     1.246e-02    2.271e-03    1.266e-02
  model.layers.5.linear_attn.in_proj_z                       81      0     1.391e-02    1.909e-03    1.404e-02
  model.layers.5.linear_attn.out_proj                        81      0     1.771e-02    2.206e-03    1.785e-02
  model.layers.5.mlp.down_proj                               81      0     1.528e-02    2.518e-03    1.548e-02
  model.layers.5.mlp.gate_proj                               81      0     1.018e-02    3.535e-03    1.077e-02
  model.layers.5.mlp.up_proj                                 81      0     1.129e-02    3.621e-03    1.185e-02
  model.layers.50.linear_attn.in_proj_qkv                    81      0     1.167e-02    2.256e-03    1.188e-02
  model.layers.50.linear_attn.in_proj_z                      81      0     1.087e-02    1.486e-03    1.097e-02
  model.layers.50.linear_attn.out_proj                       81      0     1.808e-02    2.348e-03    1.823e-02
  model.layers.50.mlp.down_proj                              81      0     1.518e-02    2.665e-03    1.541e-02
  model.layers.50.mlp.gate_proj                              81      0     9.856e-03    3.283e-03    1.039e-02
  model.layers.50.mlp.up_proj                                81      0     1.087e-02    3.439e-03    1.140e-02
  model.layers.51.mlp.down_proj                              81      0     1.525e-02    2.653e-03    1.548e-02
  model.layers.51.mlp.gate_proj                              81      0     8.703e-03    3.392e-03    9.342e-03
  model.layers.51.mlp.up_proj                                81      0     1.112e-02    3.553e-03    1.167e-02
  model.layers.51.self_attn.k_proj                           81      0     1.104e-02    2.409e-03    1.130e-02
  model.layers.51.self_attn.o_proj                           81      0     1.923e-02    2.297e-03    1.937e-02
  model.layers.51.self_attn.q_proj                           81      0     9.125e-03    2.022e-03    9.347e-03
  model.layers.51.self_attn.v_proj                           81      0     1.025e-02    2.269e-03    1.050e-02
  model.layers.52.linear_attn.in_proj_qkv                    81      0     9.626e-03    2.112e-03    9.855e-03
  model.layers.52.linear_attn.in_proj_z                      81      0     8.540e-03    1.286e-03    8.638e-03
  model.layers.52.linear_attn.out_proj                       81      0     1.872e-02    2.304e-03    1.886e-02
  model.layers.52.mlp.down_proj                              81      0     1.783e-02    2.683e-03    1.803e-02
  model.layers.52.mlp.gate_proj                              81      0     8.150e-03    3.400e-03    8.832e-03
  model.layers.52.mlp.up_proj                                81      0     1.107e-02    3.565e-03    1.163e-02
  model.layers.53.linear_attn.in_proj_qkv                    81      0     9.506e-03    2.129e-03    9.742e-03
  model.layers.53.linear_attn.in_proj_z                      81      0     8.685e-03    1.312e-03    8.785e-03
  model.layers.53.linear_attn.out_proj                       81      0     1.688e-02    2.284e-03    1.704e-02
  model.layers.53.mlp.down_proj                              81      0     1.846e-02    2.770e-03    1.866e-02
  model.layers.53.mlp.gate_proj                              81      0     7.914e-03    3.405e-03    8.616e-03
  model.layers.53.mlp.up_proj                                81      0     1.098e-02    3.567e-03    1.154e-02
  model.layers.54.linear_attn.in_proj_qkv                    81      0     9.930e-03    2.233e-03    1.018e-02
  model.layers.54.linear_attn.in_proj_z                      81      0     8.296e-03    1.277e-03    8.394e-03
  model.layers.54.linear_attn.out_proj                       81      0     1.746e-02    2.133e-03    1.759e-02
  model.layers.54.mlp.down_proj                              81      0     1.830e-02    2.512e-03    1.847e-02
  model.layers.54.mlp.gate_proj                              81      0     7.284e-03    3.355e-03    8.020e-03
  model.layers.54.mlp.up_proj                                81      0     9.182e-03    3.420e-03    9.799e-03
  model.layers.55.mlp.down_proj                              81      0     1.699e-02    2.628e-03    1.720e-02
  model.layers.55.mlp.gate_proj                              81      0     7.234e-03    3.388e-03    7.989e-03
  model.layers.55.mlp.up_proj                                81      0     1.132e-02    3.604e-03    1.188e-02
  model.layers.55.self_attn.k_proj                           81      0     1.087e-02    2.402e-03    1.113e-02
  model.layers.55.self_attn.o_proj                           81      0     1.732e-02    2.198e-03    1.746e-02
  model.layers.55.self_attn.q_proj                           81      0     8.896e-03    2.104e-03    9.143e-03
  model.layers.55.self_attn.v_proj                           81      0     9.491e-03    2.219e-03    9.745e-03
  model.layers.56.linear_attn.in_proj_qkv                    81      0     8.996e-03    2.092e-03    9.236e-03
  model.layers.56.linear_attn.in_proj_z                      81      0     8.713e-03    1.329e-03    8.815e-03
  model.layers.56.linear_attn.out_proj                       81      0     1.730e-02    2.108e-03    1.743e-02
  model.layers.56.mlp.down_proj                              81      0     1.973e-02    2.530e-03    1.989e-02
  model.layers.56.mlp.gate_proj                              81      0     7.323e-03    3.347e-03    8.052e-03
  model.layers.56.mlp.up_proj                                81      0     1.132e-02    3.599e-03    1.188e-02
  model.layers.57.linear_attn.in_proj_qkv                    81      0     8.642e-03    2.054e-03    8.885e-03
  model.layers.57.linear_attn.in_proj_z                      81      0     8.515e-03    1.299e-03    8.613e-03
  model.layers.57.linear_attn.out_proj                       81      0     1.590e-02    2.507e-03    1.610e-02
  model.layers.57.mlp.down_proj                              81      0     1.637e-02    2.798e-03    1.661e-02
  model.layers.57.mlp.gate_proj                              81      0     7.134e-03    3.326e-03    7.872e-03
  model.layers.57.mlp.up_proj                                81      0     1.116e-02    3.599e-03    1.173e-02
  model.layers.58.linear_attn.in_proj_qkv                    81      0     9.192e-03    2.133e-03    9.437e-03
  model.layers.58.linear_attn.in_proj_z                      81      0     8.763e-03    1.313e-03    8.861e-03
  model.layers.58.linear_attn.out_proj                       81      0     1.892e-02    2.407e-03    1.908e-02
  model.layers.58.mlp.down_proj                              81      0     1.657e-02    2.620e-03    1.677e-02
  model.layers.58.mlp.gate_proj                              81      0     7.016e-03    3.317e-03    7.760e-03
  model.layers.58.mlp.up_proj                                81      0     1.044e-02    3.550e-03    1.103e-02
  model.layers.59.mlp.down_proj                              81      0     1.664e-02    2.527e-03    1.683e-02
  model.layers.59.mlp.gate_proj                              81      0     6.606e-03    3.308e-03    7.388e-03
  model.layers.59.mlp.up_proj                                81      0     9.285e-03    3.448e-03    9.903e-03
  model.layers.59.self_attn.k_proj                           81      0     1.020e-02    2.349e-03    1.046e-02
  model.layers.59.self_attn.o_proj                           81      0     1.804e-02    2.288e-03    1.819e-02
  model.layers.59.self_attn.q_proj                           81      0     8.661e-03    2.117e-03    8.916e-03
  model.layers.59.self_attn.v_proj                           81      0     9.120e-03    2.162e-03    9.376e-03
  model.layers.6.linear_attn.in_proj_qkv                     81      0     1.264e-02    2.270e-03    1.285e-02
  model.layers.6.linear_attn.in_proj_z                       81      0     1.425e-02    1.992e-03    1.439e-02
  model.layers.6.linear_attn.out_proj                        81      0     1.683e-02    2.298e-03    1.698e-02
  model.layers.6.mlp.down_proj                               81      0     1.354e-02    2.403e-03    1.376e-02
  model.layers.6.mlp.gate_proj                               81      0     1.046e-02    3.490e-03    1.102e-02
  model.layers.6.mlp.up_proj                                 81      0     1.113e-02    3.560e-03    1.169e-02
  model.layers.60.linear_attn.in_proj_qkv                    81      0     7.948e-03    1.975e-03    8.189e-03
  model.layers.60.linear_attn.in_proj_z                      81      0     7.501e-03    1.194e-03    7.596e-03
  model.layers.60.linear_attn.out_proj                       81      0     1.692e-02    2.505e-03    1.710e-02
  model.layers.60.mlp.down_proj                              81      0     1.584e-02    2.710e-03    1.607e-02
  model.layers.60.mlp.gate_proj                              81      0     6.830e-03    3.245e-03    7.562e-03
  model.layers.60.mlp.up_proj                                81      0     9.185e-03    3.392e-03    9.790e-03
  model.layers.61.linear_attn.in_proj_qkv                    81      0     8.884e-03    2.091e-03    9.125e-03
  model.layers.61.linear_attn.in_proj_z                      81      0     7.949e-03    1.247e-03    8.046e-03
  model.layers.61.linear_attn.out_proj                       81      0     1.666e-02    2.347e-03    1.682e-02
  model.layers.61.mlp.down_proj                              81      0     1.498e-02    2.662e-03    1.522e-02
  model.layers.61.mlp.gate_proj                              81      0     6.566e-03    3.199e-03    7.304e-03
  model.layers.61.mlp.up_proj                                81      0     8.718e-03    3.320e-03    9.330e-03
  model.layers.62.linear_attn.in_proj_qkv                    81      0     8.142e-03    1.974e-03    8.379e-03
  model.layers.62.linear_attn.in_proj_z                      81      0     7.383e-03    1.219e-03    7.483e-03
  model.layers.62.linear_attn.out_proj                       81      0     1.517e-02    2.360e-03    1.535e-02
  model.layers.62.mlp.down_proj                              81      0     1.491e-02    2.530e-03    1.512e-02
  model.layers.62.mlp.gate_proj                              81      0     6.260e-03    3.133e-03    6.999e-03
  model.layers.62.mlp.up_proj                                81      0     6.868e-03    3.148e-03    7.554e-03
  model.layers.63.mlp.down_proj                              81      0     1.188e-02    2.006e-03    1.205e-02
  model.layers.63.mlp.gate_proj                              81      0     5.645e-03    3.069e-03    6.425e-03
  model.layers.63.mlp.up_proj                                81      0     4.720e-03    3.052e-03    5.620e-03
  model.layers.63.self_attn.k_proj                           81      0     1.075e-02    2.411e-03    1.101e-02
  model.layers.63.self_attn.o_proj                           81      0     1.295e-02    2.015e-03    1.310e-02
  model.layers.63.self_attn.q_proj                           81      0     7.002e-03    1.936e-03    7.265e-03
  model.layers.63.self_attn.v_proj                           81      0     4.688e-03    1.728e-03    4.996e-03
  model.layers.7.mlp.down_proj                               81      0     1.422e-02    2.582e-03    1.446e-02
  model.layers.7.mlp.gate_proj                               81      0     1.041e-02    3.565e-03    1.100e-02
  model.layers.7.mlp.up_proj                                 81      0     1.135e-02    3.628e-03    1.192e-02
  model.layers.7.self_attn.k_proj                            81      0     1.384e-02    2.716e-03    1.410e-02
  model.layers.7.self_attn.o_proj                            81      0     1.326e-02    2.271e-03    1.346e-02
  model.layers.7.self_attn.q_proj                            81      0     1.250e-02    2.077e-03    1.267e-02
  model.layers.7.self_attn.v_proj                            81      0     1.103e-02    2.331e-03    1.128e-02
  model.layers.8.linear_attn.in_proj_qkv                     81      0     1.244e-02    2.249e-03    1.264e-02
  model.layers.8.linear_attn.in_proj_z                       81      0     1.306e-02    1.814e-03    1.318e-02
  model.layers.8.linear_attn.out_proj                        81      0     1.616e-02    2.279e-03    1.632e-02
  model.layers.8.mlp.down_proj                               81      0     1.507e-02    2.553e-03    1.528e-02
  model.layers.8.mlp.gate_proj                               81      0     1.050e-02    3.562e-03    1.109e-02
  model.layers.8.mlp.up_proj                                 81      0     1.129e-02    3.616e-03    1.186e-02
  model.layers.9.linear_attn.in_proj_qkv                     81      0     1.310e-02    2.236e-03    1.329e-02
  model.layers.9.linear_attn.in_proj_z                       81      0     1.360e-02    1.892e-03    1.373e-02
  model.layers.9.linear_attn.out_proj                        81      0     1.603e-02    2.266e-03    1.619e-02
  model.layers.9.mlp.down_proj                               81      0     1.344e-02    2.469e-03    1.366e-02
  model.layers.9.mlp.gate_proj                               81      0     1.058e-02    3.571e-03    1.117e-02
  model.layers.9.mlp.up_proj                                 81      0     1.119e-02    3.608e-03    1.175e-02

ARM RECEIPTS -- what each arm actually executed, from the engine's
own per-call census.  fused/standalone is the reconstruct variant;
acc16/acc32 is the cuBLAS compute type leg B asked hgemm for.
  arm       legB calls          rows     fused  standalone     acc16     acc32   fp32pin
  REF            32000      65536000     32000           0     32000         0         0
  REF2           32000      65536000     32000           0     32000         0         0
  STD            32000      65536000         0       32000     32000         0         0
  REF32          32000      65536000     32000           0         0     32000         0
  STD32          32000      65536000         0       32000         0     32000         0
  INT8           32000      65536000         0       32000     32000         0         0
  all arms match their expected leg/accumulator signature
  prefill positions aligned: [8, 8, 8, 8, 8, 8, 8, 8, 8, 8] per prompt, identical across all 6 arms

THE TAIL over PREFILL positions only -- one per prefill chunk, so
every arm scores the SAME context. Decode positions are reported
separately and are NOT a numerics measure: once a token flips, the
arms are reading different contexts. The control proves it -- pooled
over decode, STD (which contains no int8 at all) has the same max KL
as INT8, so a pooled tail cannot attribute anything to the kernel.

arm            n        mean         p50         p95         p99         MAX     argmax flip
REF           80   0.000e+00   0.000e+00   0.000e+00   0.000e+00   0.000e+00 0/80 =   0.00%
REF2          80   0.000e+00   0.000e+00   0.000e+00   0.000e+00   0.000e+00 0/80 =   0.00%
STD           80   4.186e-03   2.592e-04   2.080e-02   1.028e-01   1.028e-01 2/80 =   2.50%
REF32         80   7.182e-03   2.572e-04   7.137e-03   4.852e-01   4.852e-01 1/80 =   1.25%
STD32         80   1.049e-02   2.884e-04   8.077e-03   7.456e-01   7.456e-01 1/80 =   1.25%
INT8          80   7.352e-02   9.120e-04   1.967e-01   2.380e+00   2.380e+00 4/80 =   5.00%

Tail attribution -- the 10 worst INT8 prefill positions, with the
control's KL at the SAME position. A tail that is positional shows
the control elevated too; a tail that is the kernel's does not.
   prompt  chunk      INT8 KL       STD KL  ratio INT8/STD
        9      2    2.380e+00    9.228e-02           25.79
        8      6    1.798e+00    5.397e-02           33.31
        0      7    1.290e+00    2.025e-02           63.69
        0      1    1.967e-01    1.028e-01            1.91
        0      6    4.997e-02    2.669e-03           18.73
        0      5    3.829e-02    2.080e-02            1.84
        8      4    1.666e-02    1.936e-03            8.61
        8      3    1.210e-02    1.996e-03            6.06
        9      6    8.412e-03    1.423e-03            5.91
        1      5    5.728e-03    3.205e-03            1.79
  Spearman rank correlation STD vs INT8 across all 80 prefill positions: 0.813

Decode positions (divergence-contaminated, shown for completeness):
arm            n        mean         MAX    greedy agree
REF           80   0.000e+00   0.000e+00 80/80 =  100.0%
REF2          80   0.000e+00   0.000e+00 80/80 =  100.0%
STD           80   9.532e-04   1.236e-02 80/80 =  100.0%
REF32         80   8.691e-04   1.546e-02 80/80 =  100.0%
STD32         80   1.093e-03   1.256e-02 80/80 =  100.0%
INT8          80   4.146e+00   3.248e+01 65/80 =   81.2%

SAMPLER-RELEVANT TAIL @ canonical (T=1.0, top_p=0.95, top_k=20), prefill positions only.
TV is the headline: under a drafter that tracks the target, speculative
decoding's expected accept rate is sum_t min(p_ref, p_arm), so TV IS the
expected accept-rate loss.
arm            n  setOverlap    TV mean     TV p95     TV max  truncKL p95  bndryChurn  drawDiff
REF           80     100.00%  0.000e+00  0.000e+00  0.000e+00    0.000e+00 0/ 80     0.00%
REF2          80     100.00%  0.000e+00  0.000e+00  0.000e+00    0.000e+00 0/ 80     0.00%
STD           80      98.79%  1.392e-02  4.743e-02  2.118e-01          inf 17/ 80     5.16%
REF32         80      98.90%  1.348e-02  3.995e-02  3.807e-01          inf 18/ 80     2.03%
STD32         80      98.64%  1.703e-02  4.212e-02  5.362e-01          inf 19/ 80     3.91%
INT8          80      97.55%  4.540e-02  1.552e-01  8.303e-01          inf 27/ 80     8.44%

SAMPLER-RELEVANT TAIL @ owner's k~40 (T=1.0, top_p=1.0, top_k=40), prefill positions only.
TV is the headline: under a drafter that tracks the target, speculative
decoding's expected accept rate is sum_t min(p_ref, p_arm), so TV IS the
expected accept-rate loss.
arm            n  setOverlap    TV mean     TV p95     TV max  truncKL p95  bndryChurn  drawDiff
REF           80     100.00%  0.000e+00  0.000e+00  0.000e+00    0.000e+00 0/ 80     0.00%
REF2          80     100.00%  0.000e+00  0.000e+00  0.000e+00    0.000e+00 0/ 80     0.00%
STD           80      97.29%  1.395e-02  4.479e-02  2.122e-01          inf 42/ 80     6.41%
REF32         80      96.99%  1.366e-02  3.823e-02  3.831e-01          inf 50/ 80     5.47%
STD32         80      97.26%  1.665e-02  3.704e-02  5.296e-01          inf 50/ 80     5.31%
INT8          80      93.55%  4.565e-02  1.668e-01  8.100e-01          inf 65/ 80    10.62%

FULL-VOCAB numbers below, kept and labelled: they are the right
instrument for GREEDY or beam decode, where every rank matters.

arm         first-prefill-position KL       median          max
REF                         0.000e+00    0.000e+00    0.000e+00
REF2                        0.000e+00    0.000e+00    0.000e+00
STD                         2.911e-04    6.310e-05    1.187e-03
REF32                       1.661e-04    1.195e-05    7.820e-04
STD32                       1.872e-04    1.264e-05    8.996e-04
INT8                        7.816e-04    4.077e-05    4.044e-03

PAIRED TV @ canonical (T=1.0, top_p=0.95, top_k=20), prefill positions.
Each row is A scored against B -- B is that ROW's reference, not the
table's.  TV is the expected speculative accept-rate loss, so the
column is read directly in percentage points.  The interval is a
cluster bootstrap over PROMPTS; the IQR is the raw spread over
positions with no floor applied to it.

  A       vs B        n  mean TV pp           95% CI pp    p25 pp    p50 pp    p75 pp    p95 pp    max pp
  REF2    REF        80      0.0000 [  0.0000,  0.0000]    0.0000    0.0000    0.0000    0.0000    0.0000
  STD     REF        80      1.3924 [  0.7833,  2.1293]    0.0000    0.2396    1.5124    4.7426   21.1789
  STD32   REF32      80      1.1223 [  0.6914,  1.6551]    0.0000    0.0000    1.5214    4.2088   16.5443
  REF32   REF        80      1.3483 [  0.7092,  2.3478]    0.0000    0.2887    1.5313    3.9952   38.0686
  STD32   STD        80      1.9543 [  0.8221,  3.4459]    0.0000    0.1304    2.0837    5.7662   47.7078
  INT8    STD        80      4.5484 [  1.3669,  7.8161]    0.0000    0.9674    2.8238   13.0583   84.9011
  INT8    REF        80      4.5403 [  1.3359,  7.8474]    0.0000    1.0595    2.9571   15.5238   83.0319
  INT8    STD32      80      5.0282 [  1.4136,  8.8293]    0.0000    1.3120    2.9368   42.2226   83.9256

lm_head CUT -- the divergence entering the head vs leaving it,
prefill positions, relative rms.  'in' is the residual stream after
24 layers; 'out' is the centred logits.  amp = out/in: at 1.0 the
head passes the body's divergence through unchanged, above it the
head adds its own.

  A       vs B        n   rel rms IN  rel rms OUT      amp
  REF2    REF        80   0.0000e+00   0.0000e+00      nan
  STD     REF        80   3.3423e-02   3.1702e-02    0.948
  STD32   REF32      80   2.5107e-02   2.3737e-02    0.945
  REF32   REF        80   3.9362e-02   3.7331e-02    0.948
  STD32   STD        80   4.0980e-02   3.8787e-02    0.946
  INT8    STD        80   9.0746e-02   8.3796e-02    0.923
  INT8    REF        80   9.0624e-02   8.3546e-02    0.922
  INT8    STD32      80   8.7006e-02   7.9839e-02    0.918
