Qwen90M step 190000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 185000,
+  "step": 190000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "6c683f5fcca897a05f7e0948afb2a85c2f5f99d3a2df8fe3ad69849852360942",
+    "model.safetensors": "5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "267a2d850846babea151e8cb265757c9351aed7191d41054613399dd488e4d56",
-    "training_config.json": "65390fe09f12d0188f44175607f7a9d1665082a136656274cb5d054eb998e80b",
+    "trainer.pt": "003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
-    "shard_scheduler.json": "fc6e7dce18328a420ccbbadf0eec3f835f85696a4e1a04ae07e86a4342d1ba5e",
-    "coverage.json": "f74a28e021edc403e0a2077b92693ffc8ada4dd44851adc7cd4be18fc29269e7",
+    "shard_scheduler.json": "6e7020fbb929d08abf5c36bb9ec83b6165c6ce2fb599b8042b1e42ee010831ac",
+    "coverage.json": "379675965a16132527a2a833bac7de4bb49e2af467d73478c93a42753679aff0",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 185000,
+  "step": 190000,
   "stats": {
-    "climbmix/blocks": 6660000,
-    "climbmix/targets": 11131092534,
-    "climbmix/input_tokens": 11131798123,
-    "ultra_qa/blocks": 2497500,
-    "ultra_qa/targets": 4074695441,
-    "ultra_qa/input_tokens": 4074703291,
-    "cortex/blocks": 2497500,
-    "cortex/targets": 95006899,
-    "cortex/input_tokens": 3969702317,
-    "ultra_style/blocks": 2497500,
-    "ultra_style/targets": 3950298536,
-    "ultra_style/input_tokens": 3950506046,
-    "openmath/blocks": 1665000,
-    "openmath/targets": 2572479334,
-    "openmath/input_tokens": 2573336685,
-    "rewrite6/blocks": 832500,
-    "rewrite6/targets": 1364722714,
-    "rewrite6/input_tokens": 1366219165
+    "climbmix/blocks": 6840000,
+    "climbmix/targets": 11431842491,
+    "climbmix/input_tokens": 11432567043,
+    "ultra_qa/blocks": 2565000,
+    "ultra_qa/targets": 4184969180,
+    "ultra_qa/input_tokens": 4184977239,
+    "cortex/blocks": 2565000,
+    "cortex/targets": 97573355,
+    "cortex/input_tokens": 4076931537,
+    "ultra_style/blocks": 2565000,
+    "ultra_style/targets": 4057172458,
+    "ultra_style/input_tokens": 4057385373,
+    "openmath/blocks": 1710000,
+    "openmath/targets": 2642162908,
+    "openmath/input_tokens": 2643043776,
+    "rewrite6/blocks": 855000,
+    "rewrite6/targets": 1401575108,
+    "rewrite6/input_tokens": 1403112022
   },
   "sources": {
     "climbmix": {
-      "page": 912,
-      "row": 2733,
+      "page": 937,
+      "row": 1183,
       "start": {
-        "file": 44,
+        "file": 45,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
+          "row_group": 24,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 6144,
-      "documents": 3738285
+      "offset": 2048,
+      "documents": 3839135
     },
     "rewrite6": {
-      "page": 18,
-      "row": 4079,
+      "page": 33,
+      "row": 359,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 73728
+          "row": 135168
         }
       },
       "epoch": 10,
       "offset": 0,
-      "documents": 2131847
+      "documents": 2189567
     },
     "openmath": {
-      "page": 570,
-      "row": 592,
+      "page": 585,
+      "row": 2658,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 151,
-          "row": 470
+          "row_group": 212,
+          "row": 910
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2335312
+      "offset": 0,
+      "documents": 2398818
     },
     "ultra_qa": {
-      "page": 246,
-      "row": 1981,
+      "page": 253,
+      "row": 592,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 36139
+          "row_group": 8,
+          "row": 410
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1009597
+      "documents": 1036880
     },
     "ultra_style": {
-      "page": 376,
-      "row": 1396,
+      "page": 386,
+      "row": 1768,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 84695
+          "row_group": 2,
+          "row": 40788
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 1541492
+      "offset": 2048,
+      "documents": 1582824
     },
     "cortex": {
-      "page": 48431,
-      "row": 115,
+      "page": 49741,
+      "row": 11,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 6199283
+      "documents": 6366859
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 185000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6c683f5fcca897a05f7e0948afb2a85c2f5f99d3a2df8fe3ad69849852360942", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "267a2d850846babea151e8cb265757c9351aed7191d41054613399dd488e4d56", "training_config.json": "65390fe09f12d0188f44175607f7a9d1665082a136656274cb5d054eb998e80b", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "fc6e7dce18328a420ccbbadf0eec3f835f85696a4e1a04ae07e86a4342d1ba5e", "coverage.json": "f74a28e021edc403e0a2077b92693ffc8ada4dd44851adc7cd4be18fc29269e7", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000185000"}
+{"step": 190000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "6e7020fbb929d08abf5c36bb9ec83b6165c6ce2fb599b8042b1e42ee010831ac", "coverage.json": "379675965a16132527a2a833bac7de4bb49e2af467d73478c93a42753679aff0", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000190000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c683f5fcca897a05f7e0948afb2a85c2f5f99d3a2df8fe3ad69849852360942
+oid sha256:5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71
 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": 76375.43340313055,
-      "required_pages_per_second_upper_estimate": 0.006620028995617115
+      "required_token_positions_per_second": 74817.5457905918,
+      "required_pages_per_second_upper_estimate": 0.006484995245792312
     },
     {
       "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": 9546.929175391318,
-      "required_pages_per_second_upper_estimate": 0.0038084127873748677
+      "required_token_positions_per_second": 9352.193223823975,
+      "required_pages_per_second_upper_estimate": 0.003730729704732717
     },
     {
       "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": 19093.858350782637,
-      "required_pages_per_second_upper_estimate": 0.004417028089023795
+      "required_token_positions_per_second": 18704.38644764795,
+      "required_pages_per_second_upper_estimate": 0.0043269306187049715
     },
     {
       "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": 28640.787526173957,
-      "required_pages_per_second_upper_estimate": 0.0017440736937450589
+      "required_token_positions_per_second": 28056.579671471925,
+      "required_pages_per_second_upper_estimate": 0.0017084985004954352
     },
     {
       "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": 28640.787526173957,
-      "required_pages_per_second_upper_estimate": 0.002204847106539067
+      "required_token_positions_per_second": 28056.579671471925,
+      "required_pages_per_second_upper_estimate": 0.002159873168693258
     },
     {
       "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": 22,
-      "required_token_positions_per_second": 28640.787526173957,
-      "required_pages_per_second_upper_estimate": 0.3556227265253729
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 28056.579671471925,
+      "required_pages_per_second_upper_estimate": 0.3483688201878879
     }
   ],
   "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": 63,
+    "prefetch_batches": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.5055529695468408,
-    "producer_p95_seconds": 0.33400089199858485,
-    "gpu_update_seconds": 0.9653365842239262,
-    "producer_batches_per_second": 1.9780320960162956,
-    "gpu_batches_per_second": 1.0359081136492316,
-    "producer_headroom": 1.9094667470536641,
+    "producer_batch_seconds": 0.15711286308589933,
+    "producer_p95_seconds": 0.3503623950018664,
+    "gpu_update_seconds": 0.9854372957696134,
+    "producer_batches_per_second": 6.364851230884027,
+    "gpu_batches_per_second": 1.0147779105711778,
+    "producer_headroom": 6.272161784938251,
     "sustainable": true,
-    "buffer_seconds": 60.81620480610735,
+    "buffer_seconds": 60.11167504194641,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 283448.8065840012,
-    "upload_bytes_per_second": 4514531.554258477,
+    "generation_bytes_per_second": 374399.8448116255,
+    "upload_bytes_per_second": 4221736.650489915,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 60.821515542412065,
+    "estimated_full_shards_per_hour": 80.33749111424993,
     "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": 1.2709916069979954,
-      "bytes": 5895393,
-      "pages": 35,
+      "seconds": 1.3773436970004695,
+      "bytes": 2708180,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2639917810010957,
-      "bytes": 13243059,
-      "pages": 34,
+      "seconds": 1.157878524998523,
+      "bytes": 6108198,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1510198380019574,
+      "bytes": 2279083,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 19.352378830000816,
+      "bytes": 15868354,
+      "pages": 40,
       "shards": 4
     },
     {
-      "seconds": 1.182673250001244,
-      "bytes": 5984869,
-      "pages": 35,
+      "seconds": 1.1481948339969676,
+      "bytes": 6679366,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.7554075879997981,
-      "bytes": 949451,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.3883708200010005,
+      "bytes": 2398547,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.1886153279992868,
-      "bytes": 11793828,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.0771264450013405,
+      "bytes": 6009796,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.1566395750014635,
-      "bytes": 1136088,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.8690516380011104,
+      "bytes": 4841742,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.3760789849984576,
-      "bytes": 7774763,
-      "pages": 37,
+      "seconds": 1.4456373529974371,
+      "bytes": 7578448,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.7174801740002295,
-      "bytes": 11887929,
-      "pages": 34,
+      "seconds": 1.2644371399983356,
+      "bytes": 12142497,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0596165809984086,
-      "bytes": 5348616,
-      "pages": 34,
-      "shards": 2
+      "seconds": 0.8376236600015545,
+      "bytes": 2410849,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 1.3328053339973849,
-      "bytes": 7544008,
-      "pages": 30,
+      "seconds": 1.585001501996885,
+      "bytes": 10678100,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 0.952769712999725,
-      "bytes": 1151127,
-      "pages": 34,
+      "seconds": 0.8894862470006046,
+      "bytes": 957902,
+      "pages": 28,
       "shards": 1
     },
     {
-      "seconds": 1.7508555999993405,
-      "bytes": 8239169,
-      "pages": 39,
+      "seconds": 1.1445572449993051,
+      "bytes": 7733205,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.4154862130017136,
-      "bytes": 9920178,
-      "pages": 28,
-      "shards": 3
+      "seconds": 1.1386810379990493,
+      "bytes": 6058898,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.3325980400004482,
-      "bytes": 6856977,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.2615420420006558,
+      "bytes": 12864855,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.1581080529977044,
-      "bytes": 7348060,
-      "pages": 32,
-      "shards": 3
+      "seconds": 0.9415042880027613,
+      "bytes": 5881567,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.0286378179989697,
-      "bytes": 1051160,
-      "pages": 32,
+      "seconds": 0.8555704430000333,
+      "bytes": 1226801,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.0477113340020878,
-      "bytes": 9989334,
-      "pages": 36,
+      "seconds": 0.896007103001466,
+      "bytes": 7569788,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1226200919991243,
-      "bytes": 7462663,
-      "pages": 38,
+      "seconds": 1.1828385400003754,
+      "bytes": 9643641,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 0.8068668020023324,
-      "bytes": 982732,
+      "seconds": 2.2667923209992296,
+      "bytes": 913952,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.4669019489992934,
-      "bytes": 13788589,
-      "pages": 39,
+      "seconds": 1.163448106002761,
+      "bytes": 13378237,
+      "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 1.348856757998874,
-      "bytes": 5345336,
-      "pages": 30,
+      "seconds": 0.9504098300021724,
+      "bytes": 828896,
+      "pages": 27,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1922220630003721,
+      "bytes": 10349010,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.6173429520022182,
-      "bytes": 6229619,
+      "seconds": 0.9792584309980157,
+      "bytes": 2231089,
+      "pages": 24,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3213498320001236,
+      "bytes": 5953362,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.705037297000672,
-      "bytes": 7666167,
-      "pages": 37,
-      "shards": 3
+      "seconds": 0.9344346500001848,
+      "bytes": 5832415,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 0.7579026999992493,
-      "bytes": 928826,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.198589940999227,
+      "bytes": 12036533,
+      "pages": 42,
+      "shards": 4
     },
     {
-      "seconds": 1.3881743410020135,
-      "bytes": 10608461,
-      "pages": 29,
-      "shards": 3
+      "seconds": 0.8354435640030715,
+      "bytes": 5772648,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.1401566910026304,
-      "bytes": 8280901,
-      "pages": 33,
+      "seconds": 1.2891920199981541,
+      "bytes": 7891753,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1491751960020338,
-      "bytes": 6275126,
-      "pages": 35,
+      "seconds": 0.9012639180000406,
+      "bytes": 5714275,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.3218376940021699,
-      "bytes": 6023489,
-      "pages": 34,
+      "seconds": 1.098721825001121,
+      "bytes": 5875494,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0548232209985144,
-      "bytes": 2410136,
+      "seconds": 1.1078113340008713,
+      "bytes": 2283092,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1507087749996572,
-      "bytes": 10108802,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8633549010010029,
-      "bytes": 2675452,
-      "pages": 37,
+      "seconds": 0.9954401819995837,
+      "bytes": 6010708,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.1101803599995037,
-      "bytes": 11781038,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.1992983269992692,
+      "bytes": 13571269,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.124476634999155,
-      "bytes": 9801537,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.1989007740012312,
+      "bytes": 5658429,
+      "pages": 39,
+      "shards": 2
     },
     {
-      "seconds": 0.9195323169988114,
-      "bytes": 2651824,
-      "pages": 37,
+      "seconds": 1.222567180000624,
+      "bytes": 5708293,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.357327001998783,
-      "bytes": 5924306,
-      "pages": 29,
+      "seconds": 1.0141177349978534,
+      "bytes": 2660846,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 0.9212070710018452,
-      "bytes": 2573619,
-      "pages": 34,
+      "seconds": 1.5187353149995033,
+      "bytes": 5879969,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2234259870019741,
-      "bytes": 10476491,
-      "pages": 39,
+      "seconds": 1.1908320559996355,
+      "bytes": 10104452,
+      "pages": 27,
       "shards": 3
     },
     {
-      "seconds": 1.209228235999035,
-      "bytes": 11665620,
-      "pages": 34,
+      "seconds": 1.954254078998929,
+      "bytes": 8371700,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8683807099987462,
-      "bytes": 2266346,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9792534570005955,
-      "bytes": 6085076,
+      "seconds": 1.207307631000731,
+      "bytes": 6036390,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2787357830020483,
-      "bytes": 10476048,
+      "seconds": 1.257034783000563,
+      "bytes": 2415905,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.3773436970004695,
-      "bytes": 2708180,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2379260390007403,
+      "bytes": 10087351,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.157878524998523,
-      "bytes": 6108198,
-      "pages": 36,
+      "seconds": 1.0402059740008553,
+      "bytes": 6293747,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1510198380019574,
-      "bytes": 2279083,
-      "pages": 28,
+      "seconds": 1.1847501439988264,
+      "bytes": 2572744,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 19.352378830000816,
-      "bytes": 15868354,
-      "pages": 40,
+      "seconds": 1.268939194000268,
+      "bytes": 16767391,
+      "pages": 42,
       "shards": 4
     },
     {
-      "seconds": 1.1481948339969676,
-      "bytes": 6679366,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.2854500899993582,
+      "bytes": 7334524,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.3883708200010005,
-      "bytes": 2398547,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.0401804770008312,
+      "bytes": 1254870,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.0771264450013405,
-      "bytes": 6009796,
-      "pages": 32,
+      "seconds": 1.428139841002121,
+      "bytes": 6304931,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.8690516380011104,
-      "bytes": 4841742,
+      "seconds": 0.987766936996195,
+      "bytes": 2378061,
       "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.4456373529974371,
-      "bytes": 7578448,
-      "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2644371399983356,
-      "bytes": 12142497,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8376236600015545,
-      "bytes": 2410849,
-      "pages": 28,
+      "seconds": 1.0273188179999124,
+      "bytes": 9379394,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.585001501996885,
-      "bytes": 10678100,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.4798621720037772,
+      "bytes": 13649971,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 0.8894862470006046,
-      "bytes": 957902,
-      "pages": 28,
+      "seconds": 0.8202600639997399,
+      "bytes": 1256174,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.1445572449993051,
-      "bytes": 7733205,
-      "pages": 37,
+      "seconds": 1.0791088629994192,
+      "bytes": 10296860,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1386810379990493,
-      "bytes": 6058898,
-      "pages": 29,
+      "seconds": 1.2926067090011202,
+      "bytes": 2391669,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.2615420420006558,
-      "bytes": 12864855,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.033323986994219,
+      "bytes": 6001182,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 0.9415042880027613,
-      "bytes": 5881567,
-      "pages": 28,
+      "seconds": 3.9172589509980753,
+      "bytes": 5890320,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.8555704430000333,
-      "bytes": 1226801,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3426188599987654,
+      "bytes": 12956908,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 0.896007103001466,
-      "bytes": 7569788,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.2346246180022717,
+      "bytes": 6315095,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1828385400003754,
-      "bytes": 9643641,
-      "pages": 34,
-      "shards": 3
+      "seconds": 3.684220068003924,
+      "bytes": 2396656,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 2.2667923209992296,
-      "bytes": 913952,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.1850955770059954,
+      "bytes": 6553717,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.163448106002761,
-      "bytes": 13378237,
-      "pages": 41,
-      "shards": 4
+      "seconds": 1.714512132006348,
+      "bytes": 6167274,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 0.9504098300021724,
-      "bytes": 828896,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.2202410780009814,
+      "bytes": 6552255,
+      "pages": 33,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:267a2d850846babea151e8cb265757c9351aed7191d41054613399dd488e4d56
-size 721772031
+oid sha256:003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788
+size 721770879

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
-  "prefetch_low_water": 31,
+  "prefetch_batches": 61,
+  "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,