builds/minicpm-v-4.6 and minicpm-v-4.6-moe: MiniCPM-V-4.6's markup and vocabulary

Recipes minicpm-v-4.6 (dense shares, vocab_sweep 11) and minicpm-v-4.6-moe (MoE shares), rendered through openbmb/MiniCPM-V-4.6@36f34a66's own template (auto_fmt), its vocabulary sweep (248094 rows). Pool documents sharing a 13-gram with an eval file were quarantined (purge_overlap) before the final build. For the MiniCPM-V 4.7 release if its template and added tokens fingerprint as 4.6's.

builds/minicpm-v-4.6-moe/calib_longctx.manifest.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7821895c63f2b08ad1fe17712db9693ec8e18cec22cca47a6ed2755e9f4935be
+size 7413

builds/minicpm-v-4.6-moe/calib_longctx.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba713004fdbaf5d0ab7b78bf9d66a1ee49ef439712a96d1399024a506bcb2211
+size 2726359

builds/minicpm-v-4.6-moe/calib_train.manifest.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:769a469fe0bb0a9826fc3217b4552c6ee0650e12906d33ea4cd2ec59e2a6ea36
+size 1061514

builds/minicpm-v-4.6-moe/calib_train.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:203ff4f74646a173dee19f9dd77f930547dad857c64f4d15cebcad669fca41ab
+size 17769808

builds/minicpm-v-4.6-moe/manifest.json

@@ -0,0 +1,258 @@
+{
+  "build": "minicpm-v-4.6-moe",
+  "built_at": null,
+  "build_tool_version": "build.py/1.0",
+  "recipe": {
+    "path": "recipes/minicpm-v-4.6-moe.yaml",
+    "sha256": "d845a33fe96af6c1c7094ff0d9796ab0406affb7013efdd9a723ce7f219d2ce2",
+    "seed": 20261012
+  },
+  "model": {
+    "repo": "openbmb/MiniCPM-V-4.6",
+    "tokenizer": "/src/tokenizer.json",
+    "gguf": "/gguf/minicpm-v-4.7-35b-a3b-bf16.gguf",
+    "vocab_size": 248094,
+    "layers": 40,
+    "sliding_window": null,
+    "context_length": 262144,
+    "llama_cpp_commit": "972b5fcc1967"
+  },
+  "tokenizer": {
+    "path": "/src/tokenizer.json",
+    "sha256": "33861e37bb955af1e3f3061182b820f347eba2b9c2c1011c82794bf0d6e77b54",
+    "rows": 248094,
+    "pre_tokenizer": "llama4"
+  },
+  "llama_cpp": {
+    "commit": "972b5fcc1967",
+    "note": "llama-imatrix must be given --parse-special or the chat markup in calib_train.txt is tokenised as literal text (parse_special defaults to false in that tool)"
+  },
+  "calib_train": {
+    "path": "calib_train.txt",
+    "sha256": "203ff4f74646a173dee19f9dd77f930547dad857c64f4d15cebcad669fca41ab",
+    "bytes": 17769808,
+    "documents": 2934,
+    "tokens": 4835729,
+    "target_tokens": 5000000,
+    "document_separator": "\n\n",
+    "shares_requested_percent": {
+      "agentic": 22.0,
+      "code": 16.0,
+      "graphics": 4.0,
+      "longctx": 10.0,
+      "multilingual": 18.0,
+      "reasoning": 12.0,
+      "structured": 6.0,
+      "vocab_sweep": 12.0
+    },
+    "shares_actual_percent": {
+      "agentic": 22.807,
+      "code": 16.67,
+      "graphics": 4.182,
+      "longctx": 10.417,
+      "multilingual": 18.864,
+      "reasoning": 12.421,
+      "structured": 4.337,
+      "vocab_sweep": 10.303
+    },
+    "documents_by_category": {
+      "agentic": 368,
+      "code": 499,
+      "graphics": 143,
+      "longctx": 20,
+      "multilingual": 249,
+      "reasoning": 996,
+      "structured": 451,
+      "vocab_sweep": 208
+    },
+    "tokens_by_category": {
+      "agentic": 1102887,
+      "code": 806097,
+      "graphics": 202230,
+      "longctx": 503724,
+      "multilingual": 912193,
+      "reasoning": 600665,
+      "structured": 209708,
+      "vocab_sweep": 498225
+    },
+    "synthetic_tokens": 2266030,
+    "synthetic_percent": 46.86,
+    "document_tokens": {
+      "p50": 712,
+      "p90": 3845,
+      "p95": 5907,
+      "p99": 14406,
+      "max": 45225,
+      "min": 60
+    }
+  },
+  "pool_sources": {
+    "pool/_quarantine/eval-residue.jsonl": {
+      "sha256": "9e7703a3c205b425f83f17110cf7a00b7558f8ed501b45f1fcb0f4145f5f2579",
+      "bytes": 1635378,
+      "documents": 64
+    },
+    "pool/_quarantine/removed.jsonl": {
+      "sha256": "28615d8080e4b3b95a29447402e8163786c42e23e6c5175a18783a591bde13d0",
+      "bytes": 8600246,
+      "documents": 1056
+    },
+    "pool/_quarantine/untokenizable.jsonl": {
+      "sha256": "8226d1b281ce13ceab909c06383d78176b3cc5c2d3f1a49094ef808e3f220335",
+      "bytes": 88848,
+      "documents": 2
+    },
+    "pool/agentic/synthetic/conversations.jsonl": {
+      "sha256": "40478e56fdee9bf965b0e56800e3d940e772db11e856f0170d92043d86861973",
+      "bytes": 5641610,
+      "documents": 531
+    },
+    "pool/agentic/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "24e99d36930e6547e90058654b712ae50a1294a1cdc4da7b38ae688a97537dca",
+      "bytes": 948845,
+      "documents": 50
+    },
+    "pool/code/legacy-dsv4.jsonl": {
+      "sha256": "28076d09ec882624d77b30289094ab65ae8c1fb57d3dceb55011b5e81fa99916",
+      "bytes": 471958,
+      "documents": 119
+    },
+    "pool/code/repos.jsonl": {
+      "sha256": "549d67465b052d78a4277da12033840309d767cd897f7d1280df408b43bbc916",
+      "bytes": 9133710,
+      "documents": 831
+    },
+    "pool/german/agentic_de.jsonl": {
+      "sha256": "f7afba5c9872c9e69f6fc9b0e4eb72dc1e80d8eaef1c92c736636b924eb00609",
+      "bytes": 1230904,
+      "documents": 735
+    },
+    "pool/german/germanrag.jsonl": {
+      "sha256": "23a5c7e8d2cea57ab8e01d5c8b953b81e6eb7811f12dae4b59b8396ceebf27ba",
+      "bytes": 1029498,
+      "documents": 265
+    },
+    "pool/german/longctx.jsonl": {
+      "sha256": "26d9b0f152105337fad88e098dcffa7dbcae8fe3eccd8c0a35850603b485ea60",
+      "bytes": 2488298,
+      "documents": 24
+    },
+    "pool/german/oasst2.jsonl": {
+      "sha256": "8ca93fad0812efdc21ed2ce06123627e97f41345eda8939419bd5af9b131ac01",
+      "bytes": 888635,
+      "documents": 410
+    },
+    "pool/german/ultrachat_de.jsonl": {
+      "sha256": "d42ccac07f7187c1f55f4de99335be4164f3b663d4319c2c29c31f234019d280",
+      "bytes": 1698445,
+      "documents": 488
+    },
+    "pool/german/wikipedia.jsonl": {
+      "sha256": "f0771f38f5f42aee0ff88741b7f6188b8b3bad3f0927181f1038716374285494",
+      "bytes": 2291623,
+      "documents": 443
+    },
+    "pool/graphics/legacy-dsv4.jsonl": {
+      "sha256": "09e62f79a8859e2c0f275bc38b93dde63f5751660eb3e9fc8636782cb47d3689",
+      "bytes": 2068241,
+      "documents": 335
+    },
+    "pool/graphics/repos.jsonl": {
+      "sha256": "862600d1842bd530e65884d8bc4bfcfa8892ddd1e430a927f30d35651539b7d1",
+      "bytes": 6231338,
+      "documents": 827
+    },
+    "pool/longctx/legacy-dsv4.jsonl": {
+      "sha256": "1f17bf4553c25558953fd746ce3bce278e932690e02d926395f8ecf21dacc100",
+      "bytes": 591648,
+      "documents": 10
+    },
+    "pool/longctx/module-bundles.jsonl": {
+      "sha256": "767b0cb154af75a8e5023e948f39fb3cbef424c99ba643779c5bd3e289945bb5",
+      "bytes": 9837165,
+      "documents": 83
+    },
+    "pool/multilingual/legacy-dsv4-wikipedia.jsonl": {
+      "sha256": "aa0344e84c60e1c423ac543d205bad77e3c0cc475799ddcb76d4e338cbdfb266",
+      "bytes": 420903,
+      "documents": 14
+    },
+    "pool/multilingual/wikipedia.jsonl": {
+      "sha256": "0cd11f4e6a768439f37aa354c422eb980d161e9ac2a8dde856a2f61a362480d0",
+      "bytes": 12207837,
+      "documents": 687
+    },
+    "pool/reasoning/synthetic/conversations.jsonl": {
+      "sha256": "9e890645baaad2045ac87ce9a9649082ccc37bec1c3d72b1da2a5b6c92af3e2f",
+      "bytes": 6724419,
+      "documents": 3015
+    },
+    "pool/reasoning/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "e690d0acb22a427e0b17f57ee8f255a36c5550de23775a52c9117198b02e8f95",
+      "bytes": 440412,
+      "documents": 153
+    },
+    "pool/structured/legacy-dsv4.jsonl": {
+      "sha256": "e0a6537986934b5418b6d711c2712728979e1019eb6ca40885429fc068e8c6d4",
+      "bytes": 108168,
+      "documents": 19
+    },
+    "pool/structured/repos.jsonl": {
+      "sha256": "631c8fd66fa7fc1d60fe50e257385512c77e0c1d0e1a86f88a482c29e0b2f83c",
+      "bytes": 1226353,
+      "documents": 227
+    },
+    "pool/structured/synthetic/structured-in-chat.jsonl": {
+      "sha256": "435a1be5a7b8bd230105b85fbe708649fb2ddef5a65198bdd808bcff1b3d1b64",
+      "bytes": 395812,
+      "documents": 207
+    },
+    "pool/vocab_sweep/synthetic/kolibri-1-moe.jsonl": {
+      "sha256": "3d165633bb5201090c11a63671e4f993e054cc342881007eb03b29e03d89cf27",
+      "bytes": 1323571,
+      "documents": 110
+    },
+    "pool/vocab_sweep/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "e30414d83fc63d1a5bb293a50d174eb65385efa237111f7ae001f923a9498424",
+      "bytes": 745316,
+      "documents": 65
+    },
+    "pool/vocab_sweep/synthetic/minicpm-v-4.6-moe.jsonl": {
+      "sha256": "c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4",
+      "bytes": 2509124,
+      "documents": 208
+    },
+    "pool/vocab_sweep/synthetic/minicpm-v-4.6.jsonl": {
+      "sha256": "c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4",
+      "bytes": 2509124,
+      "documents": 208
+    },
+    "pool/vocab_sweep/synthetic/muse-glimmer-30b.jsonl": {
+      "sha256": "4eb612024896df3b979fa74b92cc27ddbce56af100a14611bcfccd4ea9ce92c5",
+      "bytes": 2068343,
+      "documents": 172
+    },
+    "pool/vocab_sweep/synthetic/nemotron-3.5-lightning.jsonl": {
+      "sha256": "cd73220ef1ac82132a4c2b217f77af7b2eaea2c06939394fe36650221cc22a1d",
+      "bytes": 1269096,
+      "documents": 113
+    },
+    "pool/vocab_sweep/synthetic/qwen3.8-flash-next-moe.jsonl": {
+      "sha256": "ee4657dd2394309d06d6e6a978e020d535e0f899c0b80ca2067672889a276913",
+      "bytes": 2508961,
+      "documents": 208
+    }
+  },
+  "calib_longctx": {
+    "path": "calib_longctx.txt",
+    "sha256": "ba713004fdbaf5d0ab7b78bf9d66a1ee49ef439712a96d1399024a506bcb2211",
+    "documents": 29,
+    "tokens": 767025,
+    "document_tokens": {
+      "min": 16646,
+      "p50": 27534,
+      "max": 32681
+    },
+    "contiguous": true
+  }
+}

builds/minicpm-v-4.6/calib_longctx.manifest.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:477fc3b17ecbc565cd9cae4982285e053273592dccbbf2782e1faad93bdc269c
+size 7126

builds/minicpm-v-4.6/calib_longctx.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4dfdc3649496ffc8bb147d12eacb97d743d91fc75f97e435677f1ac2fe095d2
+size 2737299

builds/minicpm-v-4.6/calib_train.manifest.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ef173e2d1e027b17a054ea9334805c03309607563d592072db25db7a477c74b
+size 1132756

builds/minicpm-v-4.6/calib_train.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90db6d3513cf585208fdbc29cd24befbae41481376d1a09422e4b089a5ae861d
+size 17821963

builds/minicpm-v-4.6/manifest.json

@@ -0,0 +1,258 @@
+{
+  "build": "minicpm-v-4.6",
+  "built_at": null,
+  "build_tool_version": "build.py/1.0",
+  "recipe": {
+    "path": "recipes/minicpm-v-4.6.yaml",
+    "sha256": "faa250b9934256523222c567bbc6d7c6766a839ed9023bfcc22b158e52c419d8",
+    "seed": 20261012
+  },
+  "model": {
+    "repo": "openbmb/MiniCPM-V-4.6",
+    "tokenizer": "/src/tokenizer.json",
+    "gguf": "/gguf/minicpm-v-4.6-bf16.gguf",
+    "vocab_size": 248094,
+    "layers": 24,
+    "sliding_window": null,
+    "context_length": 262144,
+    "llama_cpp_commit": "972b5fcc1967"
+  },
+  "tokenizer": {
+    "path": "/src/tokenizer.json",
+    "sha256": "33861e37bb955af1e3f3061182b820f347eba2b9c2c1011c82794bf0d6e77b54",
+    "rows": 248094,
+    "pre_tokenizer": "llama4"
+  },
+  "llama_cpp": {
+    "commit": "972b5fcc1967",
+    "note": "llama-imatrix must be given --parse-special or the chat markup in calib_train.txt is tokenised as literal text (parse_special defaults to false in that tool)"
+  },
+  "calib_train": {
+    "path": "calib_train.txt",
+    "sha256": "90db6d3513cf585208fdbc29cd24befbae41481376d1a09422e4b089a5ae861d",
+    "bytes": 17821963,
+    "documents": 3108,
+    "tokens": 4950217,
+    "target_tokens": 5000000,
+    "document_separator": "\n\n",
+    "shares_requested_percent": {
+      "agentic": 24.272,
+      "code": 17.476,
+      "graphics": 2.913,
+      "longctx": 11.65,
+      "multilingual": 13.592,
+      "reasoning": 14.563,
+      "structured": 4.854,
+      "vocab_sweep": 10.68
+    },
+    "shares_actual_percent": {
+      "agentic": 24.546,
+      "code": 17.739,
+      "graphics": 2.957,
+      "longctx": 12.281,
+      "multilingual": 13.751,
+      "reasoning": 14.721,
+      "structured": 3.94,
+      "vocab_sweep": 10.065
+    },
+    "documents_by_category": {
+      "agentic": 414,
+      "code": 511,
+      "graphics": 108,
+      "longctx": 24,
+      "multilingual": 185,
+      "reasoning": 1208,
+      "structured": 450,
+      "vocab_sweep": 208
+    },
+    "tokens_by_category": {
+      "agentic": 1215084,
+      "code": 878105,
+      "graphics": 146399,
+      "longctx": 607920,
+      "multilingual": 680719,
+      "reasoning": 728703,
+      "structured": 195062,
+      "vocab_sweep": 498225
+    },
+    "synthetic_tokens": 2506265,
+    "synthetic_percent": 50.629,
+    "document_tokens": {
+      "p50": 697,
+      "p90": 3599,
+      "p95": 5562,
+      "p99": 15048,
+      "max": 45225,
+      "min": 60
+    }
+  },
+  "pool_sources": {
+    "pool/_quarantine/eval-residue.jsonl": {
+      "sha256": "9e7703a3c205b425f83f17110cf7a00b7558f8ed501b45f1fcb0f4145f5f2579",
+      "bytes": 1635378,
+      "documents": 64
+    },
+    "pool/_quarantine/removed.jsonl": {
+      "sha256": "28615d8080e4b3b95a29447402e8163786c42e23e6c5175a18783a591bde13d0",
+      "bytes": 8600246,
+      "documents": 1056
+    },
+    "pool/_quarantine/untokenizable.jsonl": {
+      "sha256": "8226d1b281ce13ceab909c06383d78176b3cc5c2d3f1a49094ef808e3f220335",
+      "bytes": 88848,
+      "documents": 2
+    },
+    "pool/agentic/synthetic/conversations.jsonl": {
+      "sha256": "40478e56fdee9bf965b0e56800e3d940e772db11e856f0170d92043d86861973",
+      "bytes": 5641610,
+      "documents": 531
+    },
+    "pool/agentic/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "24e99d36930e6547e90058654b712ae50a1294a1cdc4da7b38ae688a97537dca",
+      "bytes": 948845,
+      "documents": 50
+    },
+    "pool/code/legacy-dsv4.jsonl": {
+      "sha256": "28076d09ec882624d77b30289094ab65ae8c1fb57d3dceb55011b5e81fa99916",
+      "bytes": 471958,
+      "documents": 119
+    },
+    "pool/code/repos.jsonl": {
+      "sha256": "549d67465b052d78a4277da12033840309d767cd897f7d1280df408b43bbc916",
+      "bytes": 9133710,
+      "documents": 831
+    },
+    "pool/german/agentic_de.jsonl": {
+      "sha256": "f7afba5c9872c9e69f6fc9b0e4eb72dc1e80d8eaef1c92c736636b924eb00609",
+      "bytes": 1230904,
+      "documents": 735
+    },
+    "pool/german/germanrag.jsonl": {
+      "sha256": "23a5c7e8d2cea57ab8e01d5c8b953b81e6eb7811f12dae4b59b8396ceebf27ba",
+      "bytes": 1029498,
+      "documents": 265
+    },
+    "pool/german/longctx.jsonl": {
+      "sha256": "26d9b0f152105337fad88e098dcffa7dbcae8fe3eccd8c0a35850603b485ea60",
+      "bytes": 2488298,
+      "documents": 24
+    },
+    "pool/german/oasst2.jsonl": {
+      "sha256": "8ca93fad0812efdc21ed2ce06123627e97f41345eda8939419bd5af9b131ac01",
+      "bytes": 888635,
+      "documents": 410
+    },
+    "pool/german/ultrachat_de.jsonl": {
+      "sha256": "d42ccac07f7187c1f55f4de99335be4164f3b663d4319c2c29c31f234019d280",
+      "bytes": 1698445,
+      "documents": 488
+    },
+    "pool/german/wikipedia.jsonl": {
+      "sha256": "f0771f38f5f42aee0ff88741b7f6188b8b3bad3f0927181f1038716374285494",
+      "bytes": 2291623,
+      "documents": 443
+    },
+    "pool/graphics/legacy-dsv4.jsonl": {
+      "sha256": "09e62f79a8859e2c0f275bc38b93dde63f5751660eb3e9fc8636782cb47d3689",
+      "bytes": 2068241,
+      "documents": 335
+    },
+    "pool/graphics/repos.jsonl": {
+      "sha256": "862600d1842bd530e65884d8bc4bfcfa8892ddd1e430a927f30d35651539b7d1",
+      "bytes": 6231338,
+      "documents": 827
+    },
+    "pool/longctx/legacy-dsv4.jsonl": {
+      "sha256": "1f17bf4553c25558953fd746ce3bce278e932690e02d926395f8ecf21dacc100",
+      "bytes": 591648,
+      "documents": 10
+    },
+    "pool/longctx/module-bundles.jsonl": {
+      "sha256": "767b0cb154af75a8e5023e948f39fb3cbef424c99ba643779c5bd3e289945bb5",
+      "bytes": 9837165,
+      "documents": 83
+    },
+    "pool/multilingual/legacy-dsv4-wikipedia.jsonl": {
+      "sha256": "aa0344e84c60e1c423ac543d205bad77e3c0cc475799ddcb76d4e338cbdfb266",
+      "bytes": 420903,
+      "documents": 14
+    },
+    "pool/multilingual/wikipedia.jsonl": {
+      "sha256": "0cd11f4e6a768439f37aa354c422eb980d161e9ac2a8dde856a2f61a362480d0",
+      "bytes": 12207837,
+      "documents": 687
+    },
+    "pool/reasoning/synthetic/conversations.jsonl": {
+      "sha256": "9e890645baaad2045ac87ce9a9649082ccc37bec1c3d72b1da2a5b6c92af3e2f",
+      "bytes": 6724419,
+      "documents": 3015
+    },
+    "pool/reasoning/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "e690d0acb22a427e0b17f57ee8f255a36c5550de23775a52c9117198b02e8f95",
+      "bytes": 440412,
+      "documents": 153
+    },
+    "pool/structured/legacy-dsv4.jsonl": {
+      "sha256": "e0a6537986934b5418b6d711c2712728979e1019eb6ca40885429fc068e8c6d4",
+      "bytes": 108168,
+      "documents": 19
+    },
+    "pool/structured/repos.jsonl": {
+      "sha256": "631c8fd66fa7fc1d60fe50e257385512c77e0c1d0e1a86f88a482c29e0b2f83c",
+      "bytes": 1226353,
+      "documents": 227
+    },
+    "pool/structured/synthetic/structured-in-chat.jsonl": {
+      "sha256": "435a1be5a7b8bd230105b85fbe708649fb2ddef5a65198bdd808bcff1b3d1b64",
+      "bytes": 395812,
+      "documents": 207
+    },
+    "pool/vocab_sweep/synthetic/kolibri-1-moe.jsonl": {
+      "sha256": "3d165633bb5201090c11a63671e4f993e054cc342881007eb03b29e03d89cf27",
+      "bytes": 1323571,
+      "documents": 110
+    },
+    "pool/vocab_sweep/synthetic/legacy-dsv4-rendered.jsonl": {
+      "sha256": "e30414d83fc63d1a5bb293a50d174eb65385efa237111f7ae001f923a9498424",
+      "bytes": 745316,
+      "documents": 65
+    },
+    "pool/vocab_sweep/synthetic/minicpm-v-4.6-moe.jsonl": {
+      "sha256": "c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4",
+      "bytes": 2509124,
+      "documents": 208
+    },
+    "pool/vocab_sweep/synthetic/minicpm-v-4.6.jsonl": {
+      "sha256": "c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4",
+      "bytes": 2509124,
+      "documents": 208
+    },
+    "pool/vocab_sweep/synthetic/muse-glimmer-30b.jsonl": {
+      "sha256": "4eb612024896df3b979fa74b92cc27ddbce56af100a14611bcfccd4ea9ce92c5",
+      "bytes": 2068343,
+      "documents": 172
+    },
+    "pool/vocab_sweep/synthetic/nemotron-3.5-lightning.jsonl": {
+      "sha256": "cd73220ef1ac82132a4c2b217f77af7b2eaea2c06939394fe36650221cc22a1d",
+      "bytes": 1269096,
+      "documents": 113
+    },
+    "pool/vocab_sweep/synthetic/qwen3.8-flash-next-moe.jsonl": {
+      "sha256": "ee4657dd2394309d06d6e6a978e020d535e0f899c0b80ca2067672889a276913",
+      "bytes": 2508961,
+      "documents": 208
+    }
+  },
+  "calib_longctx": {
+    "path": "calib_longctx.txt",
+    "sha256": "b4dfdc3649496ffc8bb147d12eacb97d743d91fc75f97e435677f1ac2fe095d2",
+    "documents": 28,
+    "tokens": 776904,
+    "document_tokens": {
+      "min": 16646,
+      "p50": 29239,
+      "max": 32681
+    },
+    "contiguous": true
+  }
+}

pool/_quarantine/eval-residue.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10c4ea316927721fc6d3c21731378491702a52cc095dba6d11c133e18a6f6112
-size 1590349
+oid sha256:9e7703a3c205b425f83f17110cf7a00b7558f8ed501b45f1fcb0f4145f5f2579
+size 1635378

pool/code/repos.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4024f3432de7f57a9f48442ceae1e499110fc3acd173ab1c9de754d9cfd39c4b
-size 9176959
+oid sha256:549d67465b052d78a4277da12033840309d767cd897f7d1280df408b43bbc916
+size 9133710

pool/vocab_sweep/synthetic/minicpm-v-4.6-moe.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4
+size 2509124

pool/vocab_sweep/synthetic/minicpm-v-4.6.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8bcca4781cb4b842f42e3c4c5c699effde6aca401e85f656ace5fd4946db2b4
+size 2509124

recipes/minicpm-v-4.6-moe.yaml

@@ -0,0 +1,74 @@
+# Calibration recipe with the MoE shares, MiniCPM-V-4.6's markup and vocabulary: the build for the
+# MiniCPM-V 4.7 35B-A3B (qwen3_5_moe_text) if its template and added tokens fingerprint as 4.6's.
+# Written from recipes/qwen3.8-flash-next-moe.yaml (same Qwen3.5 BPE); see minicpm-v-4.6.yaml for the
+# template notes.
+#
+# Architecture facts that matter here (the 4.7 MoE text tower, Qwen3.5-35B-A3B layout):
+#   hidden_size = 2048 (%256 = 0)
+#   moe_intermediate_size = 512 (%256 = 0), shared expert 512
+#   256 routed experts, 8 active: ~5M * 8 / 256 = ~156k tokens per expert
+#
+# Shares (they sum to 100, so they are percentages):
+#   agentic 22      tool markup is a dense token set natural text never emits
+#   code 16         the agentic evals in this class are code shaped
+#   reasoning 12    on MoE the slack goes to multilingual and vocab, which reach rarely routed experts
+#   multilingual 18 non-Latin text routes to experts English never touches
+#   longctx 10      no sliding window; long range also gets calib_longctx
+#   vocab_sweep 12  regenerated against THIS tokenizer; every id routes somewhere
+#   structured 6    JSON, YAML, TOML, SQL, as files and inside assistant turns
+#   graphics 4
+
+name: minicpm-v-4.6-moe
+
+model:
+  repo: openbmb/MiniCPM-V-4.6
+  tokenizer: /src/tokenizer.json
+  gguf: /gguf/minicpm-v-4.7-35b-a3b-bf16.gguf
+  vocab_size: 248094
+  layers: 40
+  sliding_window: null
+  context_length: 262144
+  llama_cpp_commit: '972b5fcc1967'
+
+seed: 20261012
+
+chat:
+  format: auto
+  add_bos_per_document: false
+
+calib_train:
+  target_tokens: 5000000
+  max_doc_fraction:
+    default: 0.05
+    longctx: 0.12
+  shares:
+    agentic: 22
+    code: 16
+    reasoning: 12
+    multilingual: 18
+    longctx: 10
+    vocab_sweep: 12
+    structured: 6
+    graphics: 4
+
+calib_longctx:
+  target_tokens: 750000
+  doc_tokens_min: 16384
+  doc_tokens_max: 32768
+  sources: [longctx]
+
+exclude:
+  # Anything pre-rendered with another model's markup.
+  render: [dsv4, nemotron, muse-glimmer]
+  provenance_key: excluded_from_builds
+
+notes:
+  imatrix: >
+    llama-imatrix defaults to parse_special = false. Without --parse-special the
+    chat markup in calib_train.txt is tokenised as literal punctuation and the
+    agentic and reasoning slices calibrate on text the model never sees.
+  vision: >
+    The vision markers (<|image_pad|>, <image>, <slice>, ...) are not in the
+    corpus: a text-only llama-imatrix feeds the pad's own embedding row where
+    inference puts the mmproj's output, so text calibration cannot give them
+    honest statistics. corpus_check specials runs with them excepted.

recipes/minicpm-v-4.6.yaml

@@ -0,0 +1,74 @@
+# Calibration recipe for openbmb/MiniCPM-V-4.6 (and a dense MiniCPM-V 4.7 with the same markup).
+# Written from recipes/qwen3.8-27b.yaml: the BPE vocabulary is Qwen3.5's (same vocab hash), so the
+# shares and the pool carry over; the markup and the added tokens are MiniCPM-V's, so this build
+# renders through openbmb/MiniCPM-V-4.6's own template (auto_fmt) and sweeps its 248094-row vocabulary.
+#
+# The template is Qwen3.5-4B's except for two things: images and video are bare <|image_pad|> /
+# <|video_pad|> (no <|vision_start|>), and enable_thinking defaults to false. auto_fmt passes
+# enable_thinking per document, so the recipe's thinking on/off split reaches it. The template has
+# no reasoning effort, so chat.reasoning_effort stays off.
+#
+# Architecture facts that matter here (text tower, qwen3_5_text):
+#   hidden_size = 1024 (%256 = 0)
+#   intermediate_size = 3584 (%256 = 0)
+#   24 blocks, 3 GDN : 1 full attention; tied embeddings (token_embd is 34% of the text weights
+#   and a GET_ROWS table: the vocab sweep is what reaches its rows)
+#
+# Shares: the dense ones of qwen3.8-27b (agentic 25, code 18, reasoning 15, multilingual 14,
+# longctx 12, structured 5, graphics 3) but vocab_sweep 11: at 10 of 102 the slice was 490k tokens, the
+# sweep of this tokenizer is 498k, and 3 of its 208 documents fell out (98.99% of the reachable ids).
+
+name: minicpm-v-4.6
+
+model:
+  repo: openbmb/MiniCPM-V-4.6
+  tokenizer: /src/tokenizer.json
+  gguf: /gguf/minicpm-v-4.6-bf16.gguf
+  vocab_size: 248094
+  layers: 24
+  sliding_window: null
+  context_length: 262144
+  llama_cpp_commit: '972b5fcc1967'
+
+seed: 20261012
+
+chat:
+  format: auto
+  add_bos_per_document: false
+
+calib_train:
+  target_tokens: 5000000
+  max_doc_fraction:
+    default: 0.05
+    longctx: 0.12
+  shares:
+    agentic: 25
+    code: 18
+    reasoning: 15
+    multilingual: 14
+    longctx: 12
+    vocab_sweep: 11
+    structured: 5
+    graphics: 3
+
+calib_longctx:
+  target_tokens: 750000
+  doc_tokens_min: 16384
+  doc_tokens_max: 32768
+  sources: [longctx]
+
+exclude:
+  # Anything pre-rendered with another model's markup.
+  render: [dsv4, nemotron, muse-glimmer]
+  provenance_key: excluded_from_builds
+
+notes:
+  imatrix: >
+    llama-imatrix defaults to parse_special = false. Without --parse-special the
+    chat markup in calib_train.txt is tokenised as literal punctuation and the
+    agentic and reasoning slices calibrate on text the model never sees.
+  vision: >
+    The vision markers (<|image_pad|>, <image>, <slice>, ...) are not in the
+    corpus: a text-only llama-imatrix feeds the pad's own embedding row where
+    inference puts the mmproj's output, so text calibration cannot give them
+    honest statistics. corpus_check specials runs with them excepted.