Add Qwen3.6-35B-A3B-UD-Q4_K_XL full-v1 for unsloth/Qwen3.6-35B-A3B-GGUF@9280dd353ab5 (#4)

- Add Qwen3.6-35B-A3B-UD-Q4_K_XL full-v1 for unsloth/Qwen3.6-35B-A3B-GGUF@9280dd353ab5 (2314bf0a88b32cbb203259f5f030b2e70b3e2425)

data/unsloth/Qwen3.6-35B-A3B-GGUF/9280dd353ab587157920d5bd391ada414d84e552/gguf/Qwen3.6-35B-A3B-UD-Q4_K_XL/full-v1/analysis.json

@@ -0,0 +1,59 @@
+{
+  "memory": {
+    "base_resident_bytes": 2678180352,
+    "expert_bytes": {
+      "bytes_per_expert": 76840960,
+      "kind": "uniform"
+    },
+    "expert_tensor_bytes_total": 19671285760,
+    "full_model_bytes": 22360456160,
+    "shard_file_overhead_bytes": 10990048
+  },
+  "model": {
+    "expert_count": 256,
+    "expert_used_count": 8
+  },
+  "ranking": {
+    "mass_checkpoints": [
+      {
+        "mass_fraction": 0.320624258670914,
+        "top_n": 1
+      },
+      {
+        "mass_fraction": 0.324429432772452,
+        "top_n": 2
+      },
+      {
+        "mass_fraction": 0.3318762258353961,
+        "top_n": 4
+      },
+      {
+        "mass_fraction": 0.3460204514826405,
+        "top_n": 8
+      },
+      {
+        "mass_fraction": 0.3726827485247638,
+        "top_n": 16
+      },
+      {
+        "mass_fraction": 0.42329422660075683,
+        "top_n": 32
+      },
+      {
+        "mass_fraction": 0.51767078795539,
+        "top_n": 64
+      },
+      {
+        "mass_fraction": 0.6919257754605005,
+        "top_n": 128
+      },
+      {
+        "mass_fraction": 1.0,
+        "top_n": 256
+      }
+    ],
+    "rows": 256,
+    "sha256": "sha256:5f514ec9a2a890798f6692b2606005fb84175092ef25555f8b5e1d0c625349b9"
+  },
+  "schema_version": 1
+}

data/unsloth/Qwen3.6-35B-A3B-GGUF/9280dd353ab587157920d5bd391ada414d84e552/gguf/Qwen3.6-35B-A3B-UD-Q4_K_XL/full-v1/metadata.json

@@ -0,0 +1,28 @@
+{
+  "all_files": [
+    "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf"
+  ],
+  "all_layers": true,
+  "analysis_tool": "llama-moe-analyze",
+  "analyzer_id": "full-v1",
+  "command": {
+    "analyzer_id": "full-v1",
+    "context_size": 4096,
+    "token_count": 32
+  },
+  "created_at": "2026-04-17T08:07:52+00:00",
+  "distribution_id": "Qwen3.6-35B-A3B-UD-Q4_K_XL",
+  "file_hashes": {
+    "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf": "sha256:707a55a8a4397ecde44de0c499d3e68c1ad1d240d1da65826b4949d1043f4450"
+  },
+  "format": "gguf",
+  "primary_file": "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf",
+  "prompt_count": null,
+  "prompt_set": null,
+  "ranking_path": "ranking.csv",
+  "schema_version": 1,
+  "source_repo": "unsloth/Qwen3.6-35B-A3B-GGUF",
+  "source_revision": "9280dd353ab587157920d5bd391ada414d84e552",
+  "status": "complete",
+  "token_count": 32
+}

data/unsloth/Qwen3.6-35B-A3B-GGUF/9280dd353ab587157920d5bd391ada414d84e552/gguf/Qwen3.6-35B-A3B-UD-Q4_K_XL/full-v1/ranking.csv

@@ -0,0 +1,265 @@
+# MoE expert ranking by gate mass
+# Model: /Users/jdumay/.cache/huggingface/hub/models--unsloth--Qwen3.6-35B-A3B-GGUF/snapshots/9280dd353ab587157920d5bd391ada414d84e552/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf
+# Experts: 256 (top-8)
+# Prompts: 10 x 32 tokens
+# Layers logged: all
+# Total token-layer observations: 3304225
+#
+# Format: expert_id,gate_mass,mass_pct,selection_count
+# Sorted by gate_mass descending (hottest first)
+0,12897.1,32.0625,3277738
+243,153.063,0.380518,1539
+89,150.49,0.37412,1298
+60,149.057,0.370558,1645
+224,146.013,0.362991,1489
+64,142.587,0.354475,1426
+95,140.235,0.348627,1550
+189,140.116,0.348331,1045
+229,136.135,0.338433,1522
+36,136.132,0.338425,1348
+125,134.277,0.333815,1138
+108,133.868,0.332798,1271
+254,133.33,0.33146,1197
+167,133.256,0.331275,1316
+43,132.86,0.330292,1652
+88,132.632,0.329726,1366
+137,132.245,0.328764,1186
+165,131.32,0.326463,1255
+160,129.608,0.322208,1217
+46,129.056,0.320835,1160
+35,128.984,0.320657,1510
+134,128.404,0.319215,1327
+65,127.282,0.316426,1397
+103,126.612,0.314761,1334
+191,126.241,0.313836,1438
+182,126.235,0.313821,1221
+72,125.683,0.31245,1384
+56,125.096,0.31099,1116
+204,125.043,0.31086,1170
+203,125.007,0.31077,1321
+158,125.007,0.310769,1112
+20,124.022,0.308322,1580
+248,123.246,0.306392,1154
+220,122.568,0.304705,1275
+61,122.235,0.303879,1274
+163,121.789,0.30277,1178
+221,121.181,0.301257,1050
+206,121.022,0.300863,1151
+105,120.96,0.300708,769
+201,120.941,0.30066,976
+47,120.627,0.29988,1263
+151,119.378,0.296776,1568
+127,119.272,0.296513,1033
+99,119.161,0.296236,1090
+208,119.019,0.295884,924
+219,118.918,0.295632,1047
+171,118.732,0.29517,1007
+111,118.732,0.29517,1115
+185,118.714,0.295126,1265
+251,118.705,0.295103,1041
+121,118.695,0.295078,978
+116,118.639,0.294939,1415
+87,118.281,0.294049,1160
+209,117.858,0.292998,1164
+217,117.032,0.290944,921
+107,116.513,0.289654,953
+42,116.391,0.289351,976
+239,115.628,0.287454,937
+2,115.512,0.287164,1055
+250,115.34,0.286737,1130
+130,115.325,0.286701,1245
+128,115.31,0.286663,981
+205,115.303,0.286645,1045
+28,115.267,0.286555,1104
+252,115.165,0.286301,907
+30,114.915,0.285681,798
+162,114.525,0.284711,986
+44,114.387,0.284369,1270
+138,114.295,0.284139,970
+32,114.198,0.283898,1098
+225,114.078,0.283601,793
+41,113.612,0.282441,683
+154,112.79,0.280397,950
+176,112.788,0.280392,967
+69,112.288,0.279151,791
+37,112.096,0.278674,897
+26,112.058,0.278579,1135
+86,111.989,0.278406,1013
+202,111.718,0.277733,1047
+63,111.708,0.277709,1091
+247,111.689,0.277661,872
+218,111.589,0.277412,1028
+112,111.27,0.276619,838
+169,111.268,0.276614,984
+13,111.201,0.276447,1305
+133,110.741,0.275304,951
+22,110.69,0.275178,1174
+91,110.66,0.275102,863
+173,110.387,0.274423,876
+39,110.049,0.273584,996
+124,110.037,0.273554,1042
+98,110.025,0.273525,902
+106,109.999,0.27346,642
+233,109.938,0.273307,1053
+70,109.569,0.27239,949
+166,109.373,0.271903,944
+180,109.137,0.271316,840
+51,108.977,0.27092,1057
+198,108.871,0.270656,685
+177,108.763,0.270386,663
+238,108.697,0.270223,1056
+196,108.531,0.269811,923
+92,108.279,0.269184,1172
+215,108.18,0.268938,785
+58,108.128,0.268809,1041
+79,108.052,0.268618,836
+15,108.027,0.268558,1011
+110,108.011,0.268517,972
+11,107.797,0.267986,938
+172,107.401,0.267001,914
+40,107.34,0.26685,837
+153,107.159,0.266398,890
+143,106.979,0.265951,764
+242,106.872,0.265685,932
+144,106.858,0.265651,813
+3,106.841,0.26561,1008
+75,106.839,0.265603,952
+54,106.475,0.264698,922
+132,106.44,0.264612,938
+17,106.359,0.26441,598
+199,106.254,0.264148,899
+192,105.588,0.262493,772
+59,105.534,0.262359,945
+1,105.412,0.262055,1279
+214,105.398,0.26202,682
+194,105.087,0.261248,779
+184,105.021,0.261084,527
+255,104.999,0.261029,811
+78,104.945,0.260895,758
+84,104.917,0.260825,826
+27,104.76,0.260434,882
+71,104.571,0.259965,930
+52,104.5,0.259789,736
+175,104.442,0.259645,679
+200,104.317,0.259333,744
+195,104.209,0.259065,822
+10,104.207,0.259059,729
+152,103.981,0.258498,724
+174,103.938,0.258391,704
+102,103.895,0.258284,628
+66,103.891,0.258274,877
+83,103.844,0.258158,784
+226,103.62,0.257602,628
+93,103.383,0.257011,744
+49,103.371,0.256983,551
+57,103.199,0.256553,824
+73,103.193,0.25654,702
+117,103.142,0.256413,747
+115,103.094,0.256293,760
+45,102.958,0.255955,740
+149,102.6,0.255066,681
+123,102.465,0.254729,744
+129,102.158,0.253967,793
+25,102.056,0.253713,636
+159,101.731,0.252906,611
+53,101.159,0.251482,852
+232,101.149,0.251457,689
+236,101.135,0.251424,809
+100,101.131,0.251413,513
+147,101.089,0.251309,909
+150,100.953,0.25097,570
+228,100.72,0.250392,571
+212,100.615,0.25013,708
+148,100.426,0.24966,851
+241,99.8775,0.248297,684
+155,99.8737,0.248288,461
+24,99.7311,0.247933,655
+170,99.6893,0.247829,777
+82,99.6649,0.247768,465
+68,99.5295,0.247432,579
+85,99.5103,0.247384,863
+4,99.4539,0.247244,754
+19,99.405,0.247123,761
+126,99.3764,0.247051,677
+16,99.3484,0.246982,690
+230,99.1009,0.246366,518
+14,98.9578,0.246011,790
+141,98.7835,0.245577,603
+164,98.717,0.245412,837
+77,98.7115,0.245398,595
+145,98.7103,0.245395,798
+23,98.71,0.245395,801
+12,98.1749,0.244064,710
+146,98.1143,0.243914,761
+213,98.0514,0.243757,595
+55,98.0126,0.243661,857
+31,97.9284,0.243452,503
+227,97.8974,0.243374,570
+48,97.7811,0.243085,616
+234,97.7319,0.242963,737
+237,97.6486,0.242756,584
+34,97.6462,0.24275,558
+38,97.6346,0.242721,709
+113,97.3721,0.242069,657
+140,97.2809,0.241842,445
+253,97.2683,0.241811,448
+80,97.091,0.24137,498
+97,97.083,0.24135,677
+33,96.7513,0.240525,715
+94,96.7201,0.240448,735
+231,96.4082,0.239672,580
+90,96.3466,0.239519,678
+9,96.336,0.239493,601
+156,96.1985,0.239151,661
+157,96.0247,0.238719,798
+120,95.8212,0.238213,823
+18,95.4528,0.237297,871
+6,95.2122,0.236699,571
+210,95.2075,0.236687,508
+183,95.1461,0.236535,641
+187,95.0521,0.236301,492
+104,95.0467,0.236288,647
+223,94.8627,0.23583,433
+178,94.7889,0.235647,549
+119,94.7123,0.235456,575
+142,94.4449,0.234792,669
+193,94.3387,0.234528,707
+76,94.3314,0.234509,634
+197,94.3222,0.234487,772
+122,94.2152,0.234221,799
+168,94.0656,0.233849,317
+207,94.0196,0.233734,631
+161,94.0048,0.233697,709
+101,93.9512,0.233564,639
+114,93.8643,0.233348,641
+244,93.7773,0.233132,485
+74,93.7731,0.233121,709
+188,93.155,0.231585,599
+7,92.5198,0.230006,613
+222,92.278,0.229405,727
+139,92.1443,0.229072,642
+136,92.0111,0.228741,417
+179,91.9323,0.228545,511
+246,91.8896,0.228439,537
+118,91.8888,0.228437,481
+81,91.7182,0.228013,610
+5,91.053,0.226359,633
+186,90.6356,0.225322,521
+50,90.5887,0.225205,756
+235,90.5667,0.22515,602
+29,90.5197,0.225033,527
+109,90.1993,0.224237,590
+8,90.0863,0.223956,437
+245,89.7939,0.223229,347
+67,88.9806,0.221207,465
+131,88.8688,0.220929,468
+181,88.1751,0.219205,583
+211,87.618,0.21782,566
+135,87.2305,0.216856,499
+62,86.7473,0.215655,386
+96,85.6512,0.21293,557
+21,84.3823,0.209776,551
+249,84.0732,0.209007,460
+216,83.6198,0.20788,426
+240,81.7215,0.203161,381
+190,81.3988,0.202359,461

data/unsloth/Qwen3.6-35B-A3B-GGUF/9280dd353ab587157920d5bd391ada414d84e552/gguf/Qwen3.6-35B-A3B-UD-Q4_K_XL/full-v1/run.log

@@ -0,0 +1,333 @@
+$ /Users/jdumay/code/mesh-llm-moe-fit-checks/llama.cpp/build/bin/llama-moe-analyze -m /Users/jdumay/.cache/huggingface/hub/models--unsloth--Qwen3.6-35B-A3B-GGUF/snapshots/9280dd353ab587157920d5bd391ada414d84e552/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf --all-layers --export-ranking /Users/jdumay/.cache/mesh-llm/moe-rankings/hf-unsloth--Qwen3.6-35B-A3B-GGUF-9280dd353ab587157920d5bd391ada414d84e552-Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf.csv -n 32 -c 4096 -ngl 0
+
+[stdout]
+
+[stderr]
+ggml_metal_device_init: tensor API disabled for pre-M5 and pre-A19 devices
+ggml_metal_library_init: using embedded metal library
+ggml_metal_library_init: loaded in 9.099 sec
+ggml_metal_rsets_init: creating a residency set collection (keep_alive = 180 s)
+ggml_metal_device_init: GPU name:   MTL0
+ggml_metal_device_init: GPU family: MTLGPUFamilyApple7  (1007)
+ggml_metal_device_init: GPU family: MTLGPUFamilyCommon3 (3003)
+ggml_metal_device_init: GPU family: MTLGPUFamilyMetal4  (5002)
+ggml_metal_device_init: simdgroup reduction   = true
+ggml_metal_device_init: simdgroup matrix mul. = true
+ggml_metal_device_init: has unified memory    = true
+ggml_metal_device_init: has bfloat            = true
+ggml_metal_device_init: has tensor            = false
+ggml_metal_device_init: use residency sets    = true
+ggml_metal_device_init: use shared buffers    = true
+ggml_metal_device_init: recommendedMaxWorkingSetSize  = 115448.73 MB
+common_init_result: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on
+llama_params_fit_impl: projected to use 16 MiB of device memory vs. 110100 MiB of free device memory
+llama_params_fit_impl: will leave 110084 >= 1024 MiB of free device memory, no changes needed
+llama_params_fit: successfully fit params to free device memory
+llama_params_fit: fitting params to free memory took 0.32 seconds
+llama_model_load_from_file_impl: using device MTL0 (Apple M1 Ultra) (unknown id) - 110100 MiB free
+llama_model_loader: loaded meta data with 54 key-value pairs and 733 tensors from /Users/jdumay/.cache/huggingface/hub/models--unsloth--Qwen3.6-35B-A3B-GGUF/snapshots/9280dd353ab587157920d5bd391ada414d84e552/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf (version GGUF V3 (latest))
+llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
+llama_model_loader: - kv   0:                       general.architecture str              = qwen35moe
+llama_model_loader: - kv   1:                               general.type str              = model
+llama_model_loader: - kv   2:                     general.sampling.top_k i32              = 20
+llama_model_loader: - kv   3:                     general.sampling.top_p f32              = 0.950000
+llama_model_loader: - kv   4:                      general.sampling.temp f32              = 1.000000
+llama_model_loader: - kv   5:                               general.name str              = Qwen3.6-35B-A3B
+llama_model_loader: - kv   6:                           general.basename str              = Qwen3.6-35B-A3B
+llama_model_loader: - kv   7:                       general.quantized_by str              = Unsloth
+llama_model_loader: - kv   8:                         general.size_label str              = 35B-A3B
+llama_model_loader: - kv   9:                            general.license str              = apache-2.0
+llama_model_loader: - kv  10:                       general.license.link str              = https://huggingface.co/Qwen/Qwen3.6-3...
+llama_model_loader: - kv  11:                           general.repo_url str              = https://huggingface.co/unsloth
+llama_model_loader: - kv  12:                   general.base_model.count u32              = 1
+llama_model_loader: - kv  13:                  general.base_model.0.name str              = Qwen3.6 35B A3B
+llama_model_loader: - kv  14:          general.base_model.0.organization str              = Qwen
+llama_model_loader: - kv  15:              general.base_model.0.repo_url str              = https://huggingface.co/Qwen/Qwen3.6-3...
+llama_model_loader: - kv  16:                               general.tags arr[str,3]       = ["qwen3_5_moe", "qwen", "image-text-t...
+llama_model_loader: - kv  17:                      qwen35moe.block_count u32              = 40
+llama_model_loader: - kv  18:                   qwen35moe.context_length u32              = 262144
+llama_model_loader: - kv  19:                 qwen35moe.embedding_length u32              = 2048
+llama_model_loader: - kv  20:             qwen35moe.attention.head_count u32              = 16
+llama_model_loader: - kv  21:          qwen35moe.attention.head_count_kv u32              = 2
+llama_model_loader: - kv  22:          qwen35moe.rope.dimension_sections arr[i32,4]       = [11, 11, 10, 0]
+llama_model_loader: - kv  23:                   qwen35moe.rope.freq_base f32              = 10000000.000000
+llama_model_loader: - kv  24: qwen35moe.attention.layer_norm_rms_epsilon f32              = 0.000001
+llama_model_loader: - kv  25:                     qwen35moe.expert_count u32              = 256
+llama_model_loader: - kv  26:                qwen35moe.expert_used_count u32              = 8
+llama_model_loader: - kv  27:             qwen35moe.attention.key_length u32              = 256
+llama_model_loader: - kv  28:           qwen35moe.attention.value_length u32              = 256
+llama_model_loader: - kv  29:       qwen35moe.expert_feed_forward_length u32              = 512
+llama_model_loader: - kv  30: qwen35moe.expert_shared_feed_forward_length u32              = 512
+llama_model_loader: - kv  31:                  qwen35moe.ssm.conv_kernel u32              = 4
+llama_model_loader: - kv  32:                   qwen35moe.ssm.state_size u32              = 128
+llama_model_loader: - kv  33:                  qwen35moe.ssm.group_count u32              = 16
+llama_model_loader: - kv  34:               qwen35moe.ssm.time_step_rank u32              = 32
+llama_model_loader: - kv  35:                   qwen35moe.ssm.inner_size u32              = 4096
+llama_model_loader: - kv  36:          qwen35moe.full_attention_interval u32              = 4
+llama_model_loader: - kv  37:             qwen35moe.rope.dimension_count u32              = 64
+llama_model_loader: - kv  38:                       tokenizer.ggml.model str              = gpt2
+llama_model_loader: - kv  39:                         tokenizer.ggml.pre str              = qwen35
+llama_model_loader: - kv  40:                      tokenizer.ggml.tokens arr[str,248320]  = ["!", "\"", "#", "$", "%", "&", "'", ...
+llama_model_loader: - kv  41:                  tokenizer.ggml.token_type arr[i32,248320]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
+llama_model_loader: - kv  42:                      tokenizer.ggml.merges arr[str,247587]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
+llama_model_loader: - kv  43:                tokenizer.ggml.eos_token_id u32              = 248046
+llama_model_loader: - kv  44:            tokenizer.ggml.padding_token_id u32              = 248055
+llama_model_loader: - kv  45:                tokenizer.ggml.bos_token_id u32              = 248044
+llama_model_loader: - kv  46:               tokenizer.ggml.add_bos_token bool             = false
+llama_model_loader: - kv  47:                    tokenizer.chat_template str              = {%- set image_count = namespace(value...
+llama_model_loader: - kv  48:               general.quantization_version u32              = 2
+llama_model_loader: - kv  49:                          general.file_type u32              = 15
+llama_model_loader: - kv  50:                      quantize.imatrix.file str              = Qwen3.6-35B-A3B-GGUF/imatrix_unsloth....
+llama_model_loader: - kv  51:                   quantize.imatrix.dataset str              = unsloth_calibration_Qwen3.6-35B-A3B.txt
+llama_model_loader: - kv  52:             quantize.imatrix.entries_count u32              = 510
+llama_model_loader: - kv  53:              quantize.imatrix.chunks_count u32              = 76
+llama_model_loader: - type  f32:  361 tensors
+llama_model_loader: - type q8_0:  252 tensors
+llama_model_loader: - type q4_K:   78 tensors
+llama_model_loader: - type q5_K:   38 tensors
+llama_model_loader: - type q6_K:    4 tensors
+print_info: file format = GGUF V3 (latest)
+print_info: file type   = Q4_K - Medium
+print_info: file size   = 20.81 GiB (5.16 BPW) 
+load: 0 unused tokens
+load: printing all EOG tokens:
+load:   - 248044 ('<|endoftext|>')
+load:   - 248046 ('<|im_end|>')
+load:   - 248063 ('<|fim_pad|>')
+load:   - 248064 ('<|repo_name|>')
+load:   - 248065 ('<|file_sep|>')
+load: special tokens cache size = 33
+load: token to piece cache size = 1.7581 MB
+print_info: arch                  = qwen35moe
+print_info: vocab_only            = 0
+print_info: no_alloc              = 0
+print_info: n_ctx_train           = 262144
+print_info: n_embd                = 2048
+print_info: n_embd_inp            = 2048
+print_info: n_layer               = 40
+print_info: n_head                = 16
+print_info: n_head_kv             = 2
+print_info: n_rot                 = 64
+print_info: n_swa                 = 0
+print_info: is_swa_any            = 0
+print_info: n_embd_head_k         = 256
+print_info: n_embd_head_v         = 256
+print_info: n_gqa                 = 8
+print_info: n_embd_k_gqa          = 512
+print_info: n_embd_v_gqa          = 512
+print_info: f_norm_eps            = 0.0e+00
+print_info: f_norm_rms_eps        = 1.0e-06
+print_info: f_clamp_kqv           = 0.0e+00
+print_info: f_max_alibi_bias      = 0.0e+00
+print_info: f_logit_scale         = 0.0e+00
+print_info: f_attn_scale          = 0.0e+00
+print_info: n_ff                  = 0
+print_info: n_expert              = 256
+print_info: n_expert_used         = 8
+print_info: n_expert_groups       = 0
+print_info: n_group_used          = 0
+print_info: causal attn           = 1
+print_info: pooling type          = -1
+print_info: rope type             = 40
+print_info: rope scaling          = linear
+print_info: freq_base_train       = 10000000.0
+print_info: freq_scale_train      = 1
+print_info: n_ctx_orig_yarn       = 262144
+print_info: rope_yarn_log_mul     = 0.0000
+print_info: rope_finetuned        = unknown
+print_info: mrope sections        = [11, 11, 10, 0]
+print_info: ssm_d_conv            = 4
+print_info: ssm_d_inner           = 4096
+print_info: ssm_d_state           = 128
+print_info: ssm_dt_rank           = 32
+print_info: ssm_n_group           = 16
+print_info: ssm_dt_b_c_rms        = 0
+print_info: model type            = 35B.A3B
+print_info: model params          = 34.66 B
+print_info: general.name          = Qwen3.6-35B-A3B
+print_info: vocab type            = BPE
+print_info: n_vocab               = 248320
+print_info: n_merges              = 247587
+print_info: BOS token             = 248044 '<|endoftext|>'
+print_info: EOS token             = 248046 '<|im_end|>'
+print_info: EOT token             = 248046 '<|im_end|>'
+print_info: PAD token             = 248055 '<|vision_pad|>'
+print_info: LF token              = 198 'Ċ'
+print_info: FIM PRE token         = 248060 '<|fim_prefix|>'
+print_info: FIM SUF token         = 248062 '<|fim_suffix|>'
+print_info: FIM MID token         = 248061 '<|fim_middle|>'
+print_info: FIM PAD token         = 248063 '<|fim_pad|>'
+print_info: FIM REP token         = 248064 '<|repo_name|>'
+print_info: FIM SEP token         = 248065 '<|file_sep|>'
+print_info: EOG token             = 248044 '<|endoftext|>'
+print_info: EOG token             = 248046 '<|im_end|>'
+print_info: EOG token             = 248063 '<|fim_pad|>'
+print_info: EOG token             = 248064 '<|repo_name|>'
+print_info: EOG token             = 248065 '<|file_sep|>'
+print_info: max token length      = 256
+load_tensors: loading model tensors, this can take a while... (mmap = true, direct_io = false)
+load_tensors: offloading 0 repeating layers to GPU
+load_tensors: offloaded 0/41 layers to GPU
+load_tensors:   CPU_Mapped model buffer size = 20798.80 MiB
+load_tensors:   CPU_REPACK model buffer size = 20699.06 MiB
+.................................................................................................
+common_init_result: added <|endoftext|> logit bias = -inf
+common_init_result: added <|im_end|> logit bias = -inf
+common_init_result: added <|fim_pad|> logit bias = -inf
+common_init_result: added <|repo_name|> logit bias = -inf
+common_init_result: added <|file_sep|> logit bias = -inf
+llama_context: constructing llama_context
+llama_context: n_seq_max     = 1
+llama_context: n_ctx         = 4096
+llama_context: n_ctx_seq     = 4096
+llama_context: n_batch       = 2048
+llama_context: n_ubatch      = 512
+llama_context: causal_attn   = 1
+llama_context: flash_attn    = auto
+llama_context: kv_unified    = false
+llama_context: freq_base     = 10000000.0
+llama_context: freq_scale    = 1
+llama_context: n_ctx_seq (4096) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
+ggml_metal_init: allocating
+ggml_metal_init: found device: Apple M1 Ultra
+ggml_metal_init: picking default device: Apple M1 Ultra
+ggml_metal_init: use fusion         = true
+ggml_metal_init: use concurrency    = true
+ggml_metal_init: use graph optimize = true
+llama_context:        CPU  output buffer size =     0.95 MiB
+llama_kv_cache:        CPU KV buffer size =    80.00 MiB
+llama_kv_cache: size =   80.00 MiB (  4096 cells,  10 layers,  1/1 seqs), K (f16):   40.00 MiB, V (f16):   40.00 MiB
+llama_kv_cache: attn_rot_k = 0, n_embd_head_k_all = 256
+llama_kv_cache: attn_rot_v = 0, n_embd_head_k_all = 256
+llama_memory_recurrent:        CPU RS buffer size =    62.81 MiB
+llama_memory_recurrent: size =   62.81 MiB (     1 cells,  40 layers,  1 seqs), R (f32):    2.81 MiB, S (f32):   60.00 MiB
+sched_reserve: reserving ...
+sched_reserve: Flash Attention was auto, set to enabled
+sched_reserve: resolving fused Gated Delta Net support:
+sched_reserve: fused Gated Delta Net (autoregressive) enabled
+sched_reserve: fused Gated Delta Net (chunked) enabled
+sched_reserve:       MTL0 compute buffer size =    16.01 MiB
+sched_reserve:        CPU compute buffer size =   493.00 MiB
+sched_reserve: graph nodes  = 3729
+sched_reserve: graph splits = 281 (with bs=512), 1 (with bs=1)
+sched_reserve: reserve took 7.47 ms, sched copies = 1
+
+=== MoE Expert Routing Analysis ===
+Model experts: 256, used per token: 8
+Logging ALL MoE layers (--all-layers)
+Will export expert ranking to: /Users/jdumay/.cache/mesh-llm/moe-rankings/hf-unsloth--Qwen3.6-35B-A3B-GGUF-9280dd353ab587157920d5bd391ada414d84e552-Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf.csv
+Running 10 prompts, generating 32 tokens each
+Logging first 9999 MoE layers per eval
+
+Prompt 1/10: <|im_start|>user
+Write a Python function to find the nth Fib...
+  collected 2640 layer snapshots (total: 2640)
+Prompt 2/10: <|im_start|>user
+Write a Rust function that reads a CSV file...
+  collected 2640 layer snapshots (total: 5280)
+Prompt 3/10: <|im_start|>user
+Explain how a B-tree index works in a datab...
+  collected 2640 layer snapshots (total: 7920)
+Prompt 4/10: <|im_start|>user
+If all roses are flowers and some flowers f...
+  collected 2640 layer snapshots (total: 10560)
+Prompt 5/10: <|im_start|>user
+A train travels 120km in 2 hours. It then s...
+  collected 2640 layer snapshots (total: 13200)
+Prompt 6/10: <|im_start|>user
+Hello! What's the best way to learn a new l...
+  collected 2640 layer snapshots (total: 15840)
+Prompt 7/10: <|im_start|>user
+Tell me a joke about programmers.<|im_end|>...
+  collected 1840 layer snapshots (total: 17680)
+Prompt 8/10: <|im_start|>user
+Summarize the key differences between TCP a...
+  collected 2640 layer snapshots (total: 20320)
+Prompt 9/10: <|im_start|>user
+Translate 'The weather is beautiful today' ...
+  collected 2640 layer snapshots (total: 22960)
+Prompt 10/10: <|im_start|>user
+List 5 healthy breakfast options with brief...
+  collected 2640 layer snapshots (total: 25600)
+
+=== Expert Popularity (gate mass, summed across all tokens & logged layers) ===
+Total tokens × layers: 3304225
+
+Top 20 experts by gate mass:
+  Expert    Mass          Mass%     Selected  
+  0         12897.1294    32.06     3277738   
+  243       153.0632      0.38      1539      
+  89        150.4899      0.37      1298      
+  60        149.0570      0.37      1645      
+  224       146.0130      0.36      1489      
+  64        142.5874      0.35      1426      
+  95        140.2351      0.35      1550      
+  189       140.1161      0.35      1045      
+  229       136.1348      0.34      1522      
+  36        136.1316      0.34      1348      
+  125       134.2771      0.33      1138      
+  108       133.8681      0.33      1271      
+  254       133.3297      0.33      1197      
+  167       133.2555      0.33      1316      
+  43        132.8598      0.33      1652      
+  88        132.6322      0.33      1366      
+  137       132.2451      0.33      1186      
+  165       131.3198      0.33      1255      
+  160       129.6081      0.32      1217      
+  46        129.0560      0.32      1160      
+
+Concentration:
+  Top   4 experts: 33.2% of total gate mass
+  Top   8 experts: 34.6% of total gate mass
+  Top  16 experts: 37.3% of total gate mass
+  Top  32 experts: 42.3% of total gate mass
+  Top  64 experts: 51.8% of total gate mass
+
+Exported expert ranking to: /Users/jdumay/.cache/mesh-llm/moe-rankings/hf-unsloth--Qwen3.6-35B-A3B-GGUF-9280dd353ab587157920d5bd391ada414d84e552-Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf.csv
+Use with moe-split: --group-map <file generated from this ranking>
+
+
+=== Group Masking Analysis (best-group capture ratio) ===
+For each group count, what fraction of the unrestricted top-8 mass
+is captured by the best single group?
+
+  Groups    Replicas  Exp/Grp     Mean        P25         P50         P5        
+  2         0         128         0.999       1.000       1.000       1.000     
+  2         1         128         0.999       1.000       1.000       1.000     
+  2         2         128         0.999       1.000       1.000       1.000     
+  2         4         128         0.999       1.000       1.000       1.000     
+  2         8         128         0.999       1.000       1.000       1.000     
+
+  4         0         64          0.997       1.000       1.000       1.000     
+  4         1         64          0.997       1.000       1.000       1.000     
+  4         2         64          0.998       1.000       1.000       1.000     
+  4         4         64          0.998       1.000       1.000       1.000     
+  4         8         64          0.998       1.000       1.000       1.000     
+
+  8         0         32          0.997       1.000       1.000       1.000     
+  8         1         32          0.997       1.000       1.000       1.000     
+  8         2         32          0.997       1.000       1.000       1.000     
+  8         4         32          0.997       1.000       1.000       1.000     
+  8         8         32          0.997       1.000       1.000       1.000     
+
+=== Interpretation ===
+Mean close to 1.0 = masking barely hurts (best group captures most of top-k mass)
+Mean < 0.7 = significant quality risk from group restriction
+P5 close to 1.0 = even worst-case tokens are OK
+P5 < 0.5 = some tokens will be badly served by any single group
+
+=== Phase 1b: Masked Generation Quality (logprob comparison) ===
+Testing 4 groups (64 experts/group) vs baseline (all 256 experts)
+Using first 5 prompts, generating 32 tokens each
+
+  Group 0 (experts 0-63 + 2 hot replicas): avg logprob delta = -0.3751
+  Group 1 (experts 64-127 + 2 hot replicas): avg logprob delta = -0.1213
+  Group 2 (experts 128-191 + 2 hot replicas): avg logprob delta = -0.1904
+  Group 3 (experts 192-255 + 2 hot replicas): avg logprob delta = -0.3067
+
+=== Interpretation (logprob delta) ===
+Delta near 0.0 = masking barely affects generation quality
+Delta < -0.1 = noticeable quality loss
+Delta < -0.5 = significant degradation
+
+ggml_metal_free: deallocating