Qwen90M step 925000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 920000,
+  "step": 925000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "a1d8a464b8b679563f7500efb716f9d59a22a79a072405929bf11f45c409cd86",
+    "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "9074a1b7dfbc891f55311833a2e0a5e2999f3e3ed8b97828deb3b1f9278a4bb5",
-    "training_config.json": "aafacc05aaf9ea874dcfa3d02457799efaf5525cf7b211a4208541f6e989efee",
+    "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520",
+    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "51242912415ada2b93a379196357d105ee124cc9486c3b0247ddc722613002b3",
-    "coverage.json": "2aecf1e195b00bbe533b1deb869e79b6a89e0276b3ef3cc4a5db09a114422f4a",
+    "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9",
+    "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 920000,
+  "step": 925000,
   "stats": {
-    "climbmix/blocks": 33120000,
-    "climbmix/targets": 55365876709,
-    "climbmix/input_tokens": 55369385413,
-    "ultra_qa/blocks": 12420000,
-    "ultra_qa/targets": 20262232468,
-    "ultra_qa/input_tokens": 20262271394,
-    "cortex/blocks": 12420000,
-    "cortex/targets": 472268016,
-    "cortex/input_tokens": 19742522511,
-    "ultra_style/blocks": 12420000,
-    "ultra_style/targets": 19643639568,
-    "ultra_style/input_tokens": 19644669224,
-    "openmath/blocks": 8280000,
-    "openmath/targets": 12794177451,
-    "openmath/input_tokens": 12798440310,
-    "rewrite6/blocks": 4140000,
-    "rewrite6/targets": 6786906730,
-    "rewrite6/input_tokens": 6794347806
+    "climbmix/blocks": 33300000,
+    "climbmix/targets": 55667089526,
+    "climbmix/input_tokens": 55670617480,
+    "ultra_qa/blocks": 12487500,
+    "ultra_qa/targets": 20372477924,
+    "ultra_qa/input_tokens": 20372517045,
+    "cortex/blocks": 12487500,
+    "cortex/targets": 474833481,
+    "cortex/input_tokens": 19849928864,
+    "ultra_style/blocks": 12487500,
+    "ultra_style/targets": 19750266654,
+    "ultra_style/input_tokens": 19751301641,
+    "openmath/blocks": 8325000,
+    "openmath/targets": 12863641674,
+    "openmath/input_tokens": 12867927568,
+    "rewrite6/blocks": 4162500,
+    "rewrite6/targets": 6823898189,
+    "rewrite6/input_tokens": 6831379657
   },
   "sources": {
     "climbmix": {
-      "page": 4535,
-      "row": 3946,
+      "page": 4560,
+      "row": 2473,
       "start": {
-        "file": 219,
+        "file": 220,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 44,
+          "row_group": 62,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 18579306
+      "offset": 2048,
+      "documents": 18680233
     },
     "rewrite6": {
-      "page": 30,
-      "row": 2461,
+      "page": 44,
+      "row": 2622,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 122880
+          "row": 180224
         }
       },
       "epoch": 51,
-      "offset": 0,
-      "documents": 10600945
+      "offset": 4096,
+      "documents": 10658450
     },
     "openmath": {
-      "page": 2834,
-      "row": 3808,
+      "page": 2850,
+      "row": 1200,
       "start": {
         "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 255,
-          "row": 167
+          "row_group": 320,
+          "row": 703
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11611872
+      "documents": 11674800
     },
     "ultra_qa": {
-      "page": 1225,
-      "row": 2859,
+      "page": 1232,
+      "row": 1402,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 18028
+          "row": 46700
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 5020459
+      "offset": 2048,
+      "documents": 5047674
     },
     "ultra_style": {
-      "page": 1870,
-      "row": 3166,
+      "page": 1880,
+      "row": 3523,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 40351
+          "row": 81311
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 7662686
+      "offset": 0,
+      "documents": 7704003
     },
     "cortex": {
-      "page": 240869,
-      "row": 71,
+      "page": 242179,
+      "row": 111,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 30831303
+      "documents": 30999023
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 920000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a1d8a464b8b679563f7500efb716f9d59a22a79a072405929bf11f45c409cd86", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9074a1b7dfbc891f55311833a2e0a5e2999f3e3ed8b97828deb3b1f9278a4bb5", "training_config.json": "aafacc05aaf9ea874dcfa3d02457799efaf5525cf7b211a4208541f6e989efee", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "51242912415ada2b93a379196357d105ee124cc9486c3b0247ddc722613002b3", "coverage.json": "2aecf1e195b00bbe533b1deb869e79b6a89e0276b3ef3cc4a5db09a114422f4a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000920000"}
+{"step": 925000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9", "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000925000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1d8a464b8b679563f7500efb716f9d59a22a79a072405929bf11f45c409cd86
+oid sha256:776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 77554.4145090482,
-      "required_pages_per_second_upper_estimate": 0.00536539434677003
+      "required_token_positions_per_second": 76780.75659017764,
+      "required_pages_per_second_upper_estimate": 0.005311870896808872
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9694.301813631026,
-      "required_pages_per_second_upper_estimate": 0.004058801729325217
+      "required_token_positions_per_second": 9597.594573772205,
+      "required_pages_per_second_upper_estimate": 0.00401831242747314
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19388.60362726205,
-      "required_pages_per_second_upper_estimate": 0.0040381348236968005
+      "required_token_positions_per_second": 19195.18914754441,
+      "required_pages_per_second_upper_estimate": 0.003997851688254453
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29082.905440893075,
-      "required_pages_per_second_upper_estimate": 0.0017434757741561818
+      "required_token_positions_per_second": 28792.783721316613,
+      "required_pages_per_second_upper_estimate": 0.0017260834200577887
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29082.905440893075,
-      "required_pages_per_second_upper_estimate": 0.002261067975847635
+      "required_token_positions_per_second": 28792.783721316613,
+      "required_pages_per_second_upper_estimate": 0.002238512288261146
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 29082.905440893075,
-      "required_pages_per_second_upper_estimate": 0.37304428420483415
+      "required_token_positions_per_second": 28792.783721316613,
+      "required_pages_per_second_upper_estimate": 0.36932291429453973
     }
   ],
   "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": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 64,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.10195158224189527,
-    "producer_p95_seconds": 0.2959083680034382,
-    "gpu_update_seconds": 0.9506615512054214,
-    "producer_batches_per_second": 9.80857754249808,
-    "gpu_batches_per_second": 1.0518990683193385,
-    "producer_headroom": 9.324637541669885,
+    "producer_batch_seconds": 0.1442483061252915,
+    "producer_p95_seconds": 0.33357249200344086,
+    "gpu_update_seconds": 0.9602406029095034,
+    "producer_batches_per_second": 6.932490417817578,
+    "gpu_batches_per_second": 1.041405661216602,
+    "producer_headroom": 6.656858778469507,
     "sustainable": true,
-    "buffer_seconds": 60.84233927714697,
+    "buffer_seconds": 60.49515798329872,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 425194.20985491655,
-    "upload_bytes_per_second": 5008563.59762069,
+    "generation_bytes_per_second": 247072.63852710513,
+    "upload_bytes_per_second": 5145160.357073876,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 91.2367794202387,
+    "estimated_full_shards_per_hour": 53.01603667125573,
     "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": 1.7035883720018319,
-      "bytes": 12069614,
-      "pages": 41,
-      "shards": 3
-    },
-    {
-      "seconds": 1.154275818997121,
-      "bytes": 1133519,
-      "pages": 32,
-      "shards": 1
+      "seconds": 1.046063713998592,
+      "bytes": 5944086,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.6667711109985248,
-      "bytes": 11320824,
-      "pages": 31,
-      "shards": 4
+      "seconds": 1.323775843000476,
+      "bytes": 7235009,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.2526901849996648,
-      "bytes": 1223359,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.370845877998363,
+      "bytes": 11121979,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.1693511949997628,
-      "bytes": 7842479,
-      "pages": 39,
+      "seconds": 1.6951102359998913,
+      "bytes": 7495029,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 0.9436758240008203,
-      "bytes": 855631,
+      "seconds": 1.1257126100026653,
+      "bytes": 946737,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.5082043110014638,
-      "bytes": 16602529,
-      "pages": 32,
-      "shards": 4
+      "seconds": 1.3359714920006809,
+      "bytes": 10736001,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.5457586960001208,
-      "bytes": 7678717,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.2456738490000134,
+      "bytes": 7708205,
+      "pages": 43,
+      "shards": 3
     },
     {
-      "seconds": 0.9852680410003813,
-      "bytes": 917014,
-      "pages": 26,
+      "seconds": 1.1626391710015014,
+      "bytes": 1209462,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 3.7047144000025582,
-      "bytes": 6068293,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.3864067120011896,
+      "bytes": 11713877,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.0213589359991602,
-      "bytes": 2097340,
-      "pages": 21,
-      "shards": 2
+      "seconds": 1.506981171998632,
+      "bytes": 11206977,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.7385368430004746,
-      "bytes": 10700049,
-      "pages": 47,
-      "shards": 3
+      "seconds": 0.8770163509980193,
+      "bytes": 891965,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.4434581130008155,
-      "bytes": 11762868,
-      "pages": 29,
+      "seconds": 1.4873648249995313,
+      "bytes": 7797403,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.317350670000451,
-      "bytes": 2491706,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.1096691220009234,
+      "bytes": 969203,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.2309653659976902,
-      "bytes": 5725164,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.6109700509987306,
+      "bytes": 10483727,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 2.086732669002231,
-      "bytes": 12324271,
-      "pages": 40,
+      "seconds": 1.5793520259976503,
+      "bytes": 13424823,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 1.4604756749977241,
-      "bytes": 1030943,
-      "pages": 27,
+      "seconds": 0.9566230939963134,
+      "bytes": 1183559,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.2873325260006823,
-      "bytes": 5911958,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1188029909972101,
-      "bytes": 2149233,
-      "pages": 21,
-      "shards": 2
+      "seconds": 1.4667487440019613,
+      "bytes": 7634603,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.4335154450018308,
-      "bytes": 16606396,
-      "pages": 45,
-      "shards": 4
+      "seconds": 1.2097738270022091,
+      "bytes": 9815120,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 0.9105693140008952,
-      "bytes": 1064335,
-      "pages": 31,
+      "seconds": 1.1665725029961322,
+      "bytes": 1196873,
+      "pages": 37,
       "shards": 1
     },
     {
-      "seconds": 1.423822492997715,
-      "bytes": 7690011,
-      "pages": 35,
+      "seconds": 1.1840190020011505,
+      "bytes": 7418937,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.2330543320022116,
-      "bytes": 6359348,
-      "pages": 37,
+      "seconds": 0.9838692140037892,
+      "bytes": 6891211,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.232538223001029,
-      "bytes": 5896962,
-      "pages": 25,
-      "shards": 3
+      "seconds": 1.3423998879952705,
+      "bytes": 11990725,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.2248858210004983,
-      "bytes": 5966749,
-      "pages": 32,
+      "seconds": 1.0155430449958658,
+      "bytes": 5875734,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2219619630013767,
-      "bytes": 6395351,
-      "pages": 19,
-      "shards": 2
+      "seconds": 2.4478168570058187,
+      "bytes": 1176227,
+      "pages": 37,
+      "shards": 1
     },
     {
-      "seconds": 1.9872482999999193,
-      "bytes": 7546218,
-      "pages": 36,
+      "seconds": 1.676982698001666,
+      "bytes": 6919911,
+      "pages": 22,
       "shards": 3
     },
     {
-      "seconds": 1.5842436470011307,
-      "bytes": 5584785,
-      "pages": 34,
+      "seconds": 1.0402027899981476,
+      "bytes": 5156086,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.1840563730002032,
-      "bytes": 6277394,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.5196284359990386,
+      "bytes": 12048583,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.434792636999191,
-      "bytes": 7615048,
-      "pages": 31,
+      "seconds": 1.2240972300060093,
+      "bytes": 7561561,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 0.9521851900026377,
-      "bytes": 1184074,
-      "pages": 35,
-      "shards": 1
+      "seconds": 0.9214401089993771,
+      "bytes": 1231310,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.8348804400011431,
-      "bytes": 15799229,
-      "pages": 24,
+      "seconds": 1.6782647159998305,
+      "bytes": 12065171,
+      "pages": 33,
       "shards": 4
     },
     {
-      "seconds": 1.133582780999859,
-      "bytes": 2836972,
-      "pages": 41,
+      "seconds": 1.0347107170018717,
+      "bytes": 6246238,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1741885840019677,
-      "bytes": 6035868,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.040108234003128,
+      "bytes": 901723,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.2717247049986327,
-      "bytes": 7413901,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.5148740549993818,
+      "bytes": 14069210,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.122358404998522,
-      "bytes": 5658330,
-      "pages": 36,
+      "seconds": 1.4405004430009285,
+      "bytes": 5051876,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2760877049986448,
-      "bytes": 6145398,
-      "pages": 32,
+      "seconds": 1.163049670001783,
+      "bytes": 6248444,
+      "pages": 39,
       "shards": 2
     },
     {
-      "seconds": 1.2947614609984157,
-      "bytes": 8480974,
-      "pages": 38,
-      "shards": 3
+      "seconds": 3.389049104996957,
+      "bytes": 2610225,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.155970021001849,
-      "bytes": 5946411,
-      "pages": 30,
+      "seconds": 1.3562926899976446,
+      "bytes": 6176571,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.1933262150014343,
-      "bytes": 11280110,
-      "pages": 36,
-      "shards": 5
+      "seconds": 1.3079960410032072,
+      "bytes": 12000675,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.0185086079982284,
-      "bytes": 1287749,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.1082588089993806,
+      "bytes": 7050197,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.046063713998592,
-      "bytes": 5944086,
-      "pages": 36,
+      "seconds": 1.0682790589999058,
+      "bytes": 6087723,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.323775843000476,
-      "bytes": 7235009,
-      "pages": 30,
+      "seconds": 1.1598256950019277,
+      "bytes": 7463538,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.370845877998363,
-      "bytes": 11121979,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.2493024350042106,
+      "bytes": 5521564,
+      "pages": 41,
+      "shards": 2
     },
     {
-      "seconds": 1.6951102359998913,
-      "bytes": 7495029,
-      "pages": 40,
+      "seconds": 1.7268630670005223,
+      "bytes": 7581932,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.1257126100026653,
-      "bytes": 946737,
-      "pages": 27,
+      "seconds": 0.9802855540037854,
+      "bytes": 1121294,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.3359714920006809,
-      "bytes": 10736001,
+      "seconds": 2.06611223800428,
+      "bytes": 11832847,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.2456738490000134,
-      "bytes": 7708205,
-      "pages": 43,
-      "shards": 3
+      "seconds": 1.2940340749992174,
+      "bytes": 12187860,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1626391710015014,
-      "bytes": 1209462,
+      "seconds": 1.0110984060011106,
+      "bytes": 1179043,
       "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.3864067120011896,
-      "bytes": 11713877,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.417530385995633,
+      "bytes": 5826314,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 1.506981171998632,
-      "bytes": 11206977,
+      "seconds": 1.3201328729992383,
+      "bytes": 7683555,
       "pages": 38,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 0.8770163509980193,
-      "bytes": 891965,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.031134179997025,
+      "bytes": 5438180,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.4873648249995313,
-      "bytes": 7797403,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.582061741995858,
+      "bytes": 13195042,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.1096691220009234,
-      "bytes": 969203,
-      "pages": 31,
+      "seconds": 0.8400354289988172,
+      "bytes": 1240516,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.6109700509987306,
-      "bytes": 10483727,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.1406454610041692,
+      "bytes": 5969910,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.5793520259976503,
-      "bytes": 13424823,
-      "pages": 37,
+      "seconds": 1.6059057809980004,
+      "bytes": 11827384,
+      "pages": 24,
       "shards": 4
     },
     {
-      "seconds": 0.9566230939963134,
-      "bytes": 1183559,
-      "pages": 35,
+      "seconds": 1.091920383994875,
+      "bytes": 1379660,
+      "pages": 39,
       "shards": 1
     },
     {
-      "seconds": 1.4667487440019613,
-      "bytes": 7634603,
-      "pages": 32,
+      "seconds": 1.415137545001926,
+      "bytes": 7393395,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.2097738270022091,
-      "bytes": 9815120,
-      "pages": 37,
+      "seconds": 1.3102802950015757,
+      "bytes": 11967731,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1665725029961322,
-      "bytes": 1196873,
-      "pages": 37,
+      "seconds": 0.9184283650029101,
+      "bytes": 675909,
+      "pages": 20,
       "shards": 1
     },
     {
-      "seconds": 1.1840190020011505,
-      "bytes": 7418937,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.4080826390054426,
+      "bytes": 11404109,
+      "pages": 41,
+      "shards": 5
     },
     {
-      "seconds": 0.9838692140037892,
-      "bytes": 6891211,
-      "pages": 34,
+      "seconds": 1.4213992930017412,
+      "bytes": 1225275,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2654285969983903,
+      "bytes": 6147750,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3423998879952705,
-      "bytes": 11990725,
+      "seconds": 1.2639681400032714,
+      "bytes": 11134689,
       "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 1.0155430449958658,
-      "bytes": 5875734,
-      "pages": 28,
+      "seconds": 1.0753672879945952,
+      "bytes": 6680589,
+      "pages": 29,
       "shards": 2
+    },
+    {
+      "seconds": 1.2491716029981035,
+      "bytes": 7586487,
+      "pages": 41,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9074a1b7dfbc891f55311833a2e0a5e2999f3e3ed8b97828deb3b1f9278a4bb5
-size 721776767
+oid sha256:f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520
+size 721778943

training_config.json

@@ -53,8 +53,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 64,
-  "prefetch_low_water": 32,
+  "prefetch_batches": 63,
+  "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,