Qwen90M step 215000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 210000,
+  "step": 215000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "921781da103eac594d0d849dee7135488c4fb169435d08bae6486e287aaf16c7",
+    "model.safetensors": "fba9a241ccd9333dd7a1e6f8ef4e6f607cb802bac67c51d4e5ab08a9819714a1",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "1a7597fc28402f093934d8c6ac10a76d19bc865c277e196a6a417b18ff76bfba",
-    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
+    "trainer.pt": "e6f406062f59ad512e7807d3f45cba2fde70c4efea152b8334e10eaac4b3bc33",
+    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
     "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
-    "shard_scheduler.json": "12ca185f760b35fc0306bf401e7f24f0c1ef6c13f220f2c3f65d1588dbf74548",
-    "coverage.json": "0784ef6420a747808e25c92ea80f2676691796080a1e53c62ef8213031463920",
+    "shard_scheduler.json": "a39a540a474143f77dfcb67ff46b317dfd9988fd48b0e36afd807af861668444",
+    "coverage.json": "774fb0d49d25ba12bc1cc0687f04fde6ae04f629088fa242a46809dcd920b474",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 210000,
+  "step": 215000,
   "stats": {
-    "climbmix/blocks": 7560000,
-    "climbmix/targets": 12635674730,
-    "climbmix/input_tokens": 12636475801,
-    "ultra_qa/blocks": 2835000,
-    "ultra_qa/targets": 4625316379,
-    "ultra_qa/input_tokens": 4625325318,
-    "cortex/blocks": 2835000,
-    "cortex/targets": 107843015,
-    "cortex/input_tokens": 4506362531,
-    "ultra_style/blocks": 2835000,
-    "ultra_style/targets": 4484394327,
-    "ultra_style/input_tokens": 4484629586,
-    "openmath/blocks": 1890000,
-    "openmath/targets": 2920105808,
-    "openmath/input_tokens": 2921079037,
-    "rewrite6/blocks": 945000,
-    "rewrite6/targets": 1549174599,
-    "rewrite6/input_tokens": 1550873200
+    "climbmix/blocks": 7740000,
+    "climbmix/targets": 12937648715,
+    "climbmix/input_tokens": 12938468575,
+    "ultra_qa/blocks": 2902500,
+    "ultra_qa/targets": 4735458775,
+    "ultra_qa/input_tokens": 4735467933,
+    "cortex/blocks": 2902500,
+    "cortex/targets": 110418665,
+    "cortex/input_tokens": 4613613081,
+    "ultra_style/blocks": 2902500,
+    "ultra_style/targets": 4591380949,
+    "ultra_style/input_tokens": 4591621680,
+    "openmath/blocks": 1935000,
+    "openmath/targets": 2989665915,
+    "openmath/input_tokens": 2990662278,
+    "rewrite6/blocks": 967500,
+    "rewrite6/targets": 1586083890,
+    "rewrite6/input_tokens": 1587822751
   },
   "sources": {
     "climbmix": {
-      "page": 1035,
-      "row": 3928,
+      "page": 1060,
+      "row": 844,
       "start": {
-        "file": 50,
+        "file": 51,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
+          "row_group": 23,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 4243288
+      "offset": 16384,
+      "documents": 4342604
     },
     "rewrite6": {
-      "page": 39,
-      "row": 727,
+      "page": 3,
+      "row": 133,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 159744
+          "row": 12288
         }
       },
-      "epoch": 11,
-      "offset": 0,
-      "documents": 2419915
+      "epoch": 12,
+      "offset": 2048,
+      "documents": 2477269
     },
     "openmath": {
-      "page": 647,
-      "row": 820,
+      "page": 662,
+      "row": 2443,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 30,
-          "row": 212
+          "row_group": 91,
+          "row": 652
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 2650932
+      "documents": 2713995
     },
     "ultra_qa": {
-      "page": 279,
-      "row": 3264,
+      "page": 286,
+      "row": 1881,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 38970
+          "row_group": 2,
+          "row": 3058
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1146048
+      "offset": 4096,
+      "documents": 1173337
     },
     "ultra_style": {
-      "page": 427,
-      "row": 231,
+      "page": 437,
+      "row": 479,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 38971
+          "row": 79931
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1749223
+      "documents": 1790431
     },
     "cortex": {
-      "page": 54982,
-      "row": 96,
+      "page": 56290,
+      "row": 85,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 7037792
+      "documents": 7205205
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 210000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "921781da103eac594d0d849dee7135488c4fb169435d08bae6486e287aaf16c7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1a7597fc28402f093934d8c6ac10a76d19bc865c277e196a6a417b18ff76bfba", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "12ca185f760b35fc0306bf401e7f24f0c1ef6c13f220f2c3f65d1588dbf74548", "coverage.json": "0784ef6420a747808e25c92ea80f2676691796080a1e53c62ef8213031463920", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000210000"}
+{"step": 215000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "fba9a241ccd9333dd7a1e6f8ef4e6f607cb802bac67c51d4e5ab08a9819714a1", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "e6f406062f59ad512e7807d3f45cba2fde70c4efea152b8334e10eaac4b3bc33", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "a39a540a474143f77dfcb67ff46b317dfd9988fd48b0e36afd807af861668444", "coverage.json": "774fb0d49d25ba12bc1cc0687f04fde6ae04f629088fa242a46809dcd920b474", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000215000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:921781da103eac594d0d849dee7135488c4fb169435d08bae6486e287aaf16c7
+oid sha256:fba9a241ccd9333dd7a1e6f8ef4e6f607cb802bac67c51d4e5ab08a9819714a1
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 74616.38181288075,
-      "required_pages_per_second_upper_estimate": 0.006467558862049758
+      "required_token_positions_per_second": 75396.39935795023,
+      "required_pages_per_second_upper_estimate": 0.006535168805919986
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9327.047726610093,
-      "required_pages_per_second_upper_estimate": 0.0037206987899354132
+      "required_token_positions_per_second": 9424.549919743778,
+      "required_pages_per_second_upper_estimate": 0.003759593872564137
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18654.095453220187,
-      "required_pages_per_second_upper_estimate": 0.004315296682235381
+      "required_token_positions_per_second": 18849.099839487557,
+      "required_pages_per_second_upper_estimate": 0.004360407515038354
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 27981.14317983028,
-      "required_pages_per_second_upper_estimate": 0.0017039048139391429
+      "required_token_positions_per_second": 28273.649759231335,
+      "required_pages_per_second_upper_estimate": 0.00172171692996125
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 27981.14317983028,
-      "required_pages_per_second_upper_estimate": 0.00215406585874511
+      "required_token_positions_per_second": 28273.649759231335,
+      "required_pages_per_second_upper_estimate": 0.002176583824937439
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.042780299000014566,
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
-      "estimated_pages_for_buffer": 21,
-      "required_token_positions_per_second": 27981.14317983028,
-      "required_pages_per_second_upper_estimate": 0.3474321514313953
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28273.649759231335,
+      "required_pages_per_second_upper_estimate": 0.35106410419100953
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 61,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.11913276657806193,
-    "producer_p95_seconds": 0.32301659399854543,
-    "gpu_update_seconds": 0.9880940111099384,
-    "producer_batches_per_second": 8.393996284345066,
-    "gpu_batches_per_second": 1.0120494495019632,
-    "producer_headroom": 8.294057457840436,
+    "producer_batch_seconds": 0.06690109020320278,
+    "producer_p95_seconds": 0.21632020699689747,
+    "gpu_update_seconds": 0.977871630845004,
+    "producer_batches_per_second": 14.947439525464214,
+    "gpu_batches_per_second": 1.0226291145555315,
+    "producer_headroom": 14.616677065722763,
     "sustainable": true,
-    "buffer_seconds": 60.273734677706244,
+    "buffer_seconds": 60.62804111239024,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 81,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 355577.373587176,
-    "upload_bytes_per_second": 5217106.874362177,
+    "generation_bytes_per_second": 294982.2025964519,
+    "upload_bytes_per_second": 5236085.920943062,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 76.29862695418795,
+    "estimated_full_shards_per_hour": 63.29631384296577,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [
     {
-      "seconds": 0.9532060609999462,
-      "bytes": 7422959,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2087593340002059,
-      "bytes": 6148456,
-      "pages": 35,
+      "seconds": 0.9283750220001821,
+      "bytes": 2657482,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2602108099999896,
-      "bytes": 2561145,
-      "pages": 35,
-      "shards": 3
+      "seconds": 0.9979908650002471,
+      "bytes": 5962542,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.5218778089993066,
-      "bytes": 10001472,
-      "pages": 31,
+      "seconds": 1.4787909560000116,
+      "bytes": 11055954,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.655338631999257,
-      "bytes": 6858339,
-      "pages": 37,
+      "seconds": 1.0256725109993567,
+      "bytes": 2318063,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3612198569990142,
-      "bytes": 7511604,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.111224463000326,
-      "bytes": 5878363,
+      "seconds": 0.9941544879984576,
+      "bytes": 6059545,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.1176849410003342,
-      "bytes": 5784032,
-      "pages": 33,
+      "seconds": 1.1885242320004181,
+      "bytes": 7012163,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2254727769995952,
-      "bytes": 7561832,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.3779209780004749,
+      "bytes": 11061746,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.1802483909996226,
-      "bytes": 1055971,
-      "pages": 33,
+      "seconds": 0.9865021240002534,
+      "bytes": 862370,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.59867495799881,
-      "bytes": 13042114,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.1108503920004296,
+      "bytes": 7694144,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1432860509994498,
-      "bytes": 5906480,
-      "pages": 31,
+      "seconds": 2.1076598479994573,
+      "bytes": 6179293,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2417195279995212,
-      "bytes": 5623378,
-      "pages": 33,
+      "seconds": 1.0649322700010089,
+      "bytes": 5823579,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.9634915199985699,
-      "bytes": 7597611,
-      "pages": 38,
+      "seconds": 1.1941432809999242,
+      "bytes": 7137236,
+      "pages": 22,
       "shards": 3
     },
     {
-      "seconds": 0.761399758999687,
-      "bytes": 1061264,
-      "pages": 32,
-      "shards": 1
+      "seconds": 1.292083440001079,
+      "bytes": 7124472,
+      "pages": 44,
+      "shards": 2
     },
     {
-      "seconds": 1.0512442660001398,
-      "bytes": 5862459,
-      "pages": 28,
+      "seconds": 1.2287268149993906,
+      "bytes": 6045685,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2080343199995696,
-      "bytes": 17359027,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.297586668999429,
+      "bytes": 7207958,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 1.0087294929999189,
-      "bytes": 1073260,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.0006021399985912,
+      "bytes": 4975022,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.4433996169991588,
-      "bytes": 7239065,
+      "seconds": 1.1054393590002292,
+      "bytes": 5686965,
       "pages": 29,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 0.894821537000098,
-      "bytes": 1079304,
-      "pages": 31,
-      "shards": 1
+      "seconds": 0.952104601999963,
+      "bytes": 2531508,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.1567702649990679,
-      "bytes": 10579644,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.6972403340005258,
+      "bytes": 16066476,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.6480453410003975,
-      "bytes": 7697533,
-      "pages": 38,
+      "seconds": 1.3755702170001314,
+      "bytes": 7451679,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.2853153289997863,
-      "bytes": 6668992,
-      "pages": 31,
+      "seconds": 1.598954668999795,
+      "bytes": 1262837,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.0224317270003667,
+      "bytes": 6094784,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2017098229989642,
-      "bytes": 7521074,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.438195700000506,
+      "bytes": 12388203,
+      "pages": 31,
+      "shards": 4
     },
     {
-      "seconds": 1.335906720998537,
-      "bytes": 5362985,
-      "pages": 26,
-      "shards": 2
+      "seconds": 0.9514959919997636,
+      "bytes": 336067,
+      "pages": 9,
+      "shards": 1
     },
     {
-      "seconds": 1.582250990999455,
-      "bytes": 5900156,
-      "pages": 27,
+      "seconds": 1.108248578999337,
+      "bytes": 861803,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.2782940739998594,
-      "bytes": 7640268,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.207310141000562,
+      "bytes": 13710377,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 0.723643687000731,
-      "bytes": 1209843,
+      "seconds": 0.9586756399985461,
+      "bytes": 1163538,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.586552576000031,
-      "bytes": 9572867,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2363076699984958,
-      "bytes": 8177813,
-      "pages": 33,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3350522609998734,
-      "bytes": 6097996,
-      "pages": 35,
+      "seconds": 1.222915964999629,
+      "bytes": 5976864,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 0.9499315409993869,
-      "bytes": 5779343,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.3758100840004772,
+      "bytes": 11655433,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.612262433000069,
-      "bytes": 7066450,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.3469852179987356,
+      "bytes": 1264752,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.120285896000496,
-      "bytes": 6092850,
-      "pages": 35,
+      "seconds": 1.1859561970013601,
+      "bytes": 6123105,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.532518773999982,
-      "bytes": 2518854,
-      "pages": 32,
+      "seconds": 0.9817681739987165,
+      "bytes": 8356010,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.2675222210000356,
-      "bytes": 11869039,
-      "pages": 39,
+      "seconds": 1.2058971119986381,
+      "bytes": 10431329,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.6028917950006871,
-      "bytes": 10461677,
-      "pages": 31,
+      "seconds": 1.240897143001348,
+      "bytes": 7984887,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.3869329340013792,
-      "bytes": 2462257,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.0015397960014525,
+      "bytes": 971215,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.3128174869998475,
-      "bytes": 5972623,
-      "pages": 34,
+      "seconds": 1.235119819000829,
+      "bytes": 5680772,
+      "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 0.8817397439997876,
-      "bytes": 1063209,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.6997340180005267,
-      "bytes": 11573294,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.104966183000215,
+      "bytes": 6761570,
+      "pages": 43,
+      "shards": 3
     },
     {
-      "seconds": 1.4719399390014587,
-      "bytes": 11935199,
-      "pages": 35,
+      "seconds": 1.8051689160020032,
+      "bytes": 11661330,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9283750220001821,
-      "bytes": 2657482,
-      "pages": 36,
+      "seconds": 1.1013407509999524,
+      "bytes": 6091571,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9979908650002471,
-      "bytes": 5962542,
-      "pages": 27,
+      "seconds": 0.8600726500008022,
+      "bytes": 2195975,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.4787909560000116,
-      "bytes": 11055954,
-      "pages": 41,
+      "seconds": 1.1491692820018216,
+      "bytes": 11269048,
+      "pages": 45,
       "shards": 3
     },
     {
-      "seconds": 1.0256725109993567,
-      "bytes": 2318063,
-      "pages": 28,
+      "seconds": 1.0205925500013109,
+      "bytes": 2657105,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9941544879984576,
-      "bytes": 6059545,
+      "seconds": 1.0683662000010372,
+      "bytes": 6645182,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.1885242320004181,
-      "bytes": 7012163,
-      "pages": 35,
+      "seconds": 1.0938033080019522,
+      "bytes": 6719573,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.3779209780004749,
-      "bytes": 11061746,
-      "pages": 35,
+      "seconds": 1.3891484439991473,
+      "bytes": 11561860,
+      "pages": 34,
       "shards": 4
     },
     {
-      "seconds": 0.9865021240002534,
-      "bytes": 862370,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.268800834997819,
+      "bytes": 6327049,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1108503920004296,
-      "bytes": 7694144,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.941673843000899,
+      "bytes": 1008328,
+      "pages": 30,
+      "shards": 1
     },
     {
-      "seconds": 2.1076598479994573,
-      "bytes": 6179293,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.119837071000802,
+      "bytes": 7224601,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.0649322700010089,
-      "bytes": 5823579,
-      "pages": 32,
-      "shards": 2
+      "seconds": 1.2717098209977848,
+      "bytes": 11010766,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1941432809999242,
-      "bytes": 7137236,
-      "pages": 22,
+      "seconds": 1.1728442359999462,
+      "bytes": 8097014,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.292083440001079,
-      "bytes": 7124472,
-      "pages": 44,
+      "seconds": 1.3505811920003907,
+      "bytes": 5904996,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.2287268149993906,
-      "bytes": 6045685,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.9558858569980657,
+      "bytes": 1199021,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.297586668999429,
-      "bytes": 7207958,
-      "pages": 28,
-      "shards": 3
+      "seconds": 1.896848256001249,
+      "bytes": 11337881,
+      "pages": 42,
+      "shards": 4
     },
     {
-      "seconds": 1.0006021399985912,
-      "bytes": 4975022,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1246725650016742,
+      "bytes": 878351,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.1054393590002292,
-      "bytes": 5686965,
-      "pages": 29,
+      "seconds": 1.1035908939993533,
+      "bytes": 6360663,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.952104601999963,
-      "bytes": 2531508,
-      "pages": 33,
-      "shards": 2
+      "seconds": 0.9343603500019526,
+      "bytes": 2109863,
+      "pages": 20,
+      "shards": 3
     },
     {
-      "seconds": 1.6972403340005258,
-      "bytes": 16066476,
-      "pages": 41,
+      "seconds": 1.3331431920014438,
+      "bytes": 16665888,
+      "pages": 43,
       "shards": 4
     },
     {
-      "seconds": 1.3755702170001314,
-      "bytes": 7451679,
-      "pages": 30,
+      "seconds": 1.675639938999666,
+      "bytes": 7647558,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.598954668999795,
-      "bytes": 1262837,
+      "seconds": 0.968150565000542,
+      "bytes": 1141925,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.0224317270003667,
-      "bytes": 6094784,
+      "seconds": 1.0775915699996403,
+      "bytes": 5849379,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3228814079993754,
+      "bytes": 11625452,
       "pages": 35,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1623286959984398,
+      "bytes": 7130450,
+      "pages": 38,
       "shards": 2
+    },
+    {
+      "seconds": 1.6525030769989826,
+      "bytes": 8063307,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0988706719981565,
+      "bytes": 771997,
+      "pages": 25,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a7597fc28402f093934d8c6ac10a76d19bc865c277e196a6a417b18ff76bfba
-size 721768575
+oid sha256:e6f406062f59ad512e7807d3f45cba2fde70c4efea152b8334e10eaac4b3bc33
+size 721752255

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 61,
-  "prefetch_low_water": 30,
+  "prefetch_batches": 62,
+  "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,