Qwen90M step 185000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 180000,
+  "step": 185000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "68121455268768b4b43f0edce892dc1b160432a06c7d9a6bf2a59971a05dac46",
+    "model.safetensors": "6c683f5fcca897a05f7e0948afb2a85c2f5f99d3a2df8fe3ad69849852360942",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "25a0a70fc0e8c2815d5f59181512453873f858013da77228c792c3eefdaa0f65",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "267a2d850846babea151e8cb265757c9351aed7191d41054613399dd488e4d56",
+    "training_config.json": "65390fe09f12d0188f44175607f7a9d1665082a136656274cb5d054eb998e80b",
     "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
-    "shard_scheduler.json": "fa0c68dff9ecb9880754b350e4c1810c4b7973ad3c5ba843eff2dd61df715747",
-    "coverage.json": "8559bd7833f4eb51cd101d3f44a0ee6e30b948a85612ba19d9f1f4a93d4321de",
+    "shard_scheduler.json": "fc6e7dce18328a420ccbbadf0eec3f835f85696a4e1a04ae07e86a4342d1ba5e",
+    "coverage.json": "f74a28e021edc403e0a2077b92693ffc8ada4dd44851adc7cd4be18fc29269e7",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 180000,
+  "step": 185000,
   "stats": {
-    "climbmix/blocks": 6480000,
-    "climbmix/targets": 10830284056,
-    "climbmix/input_tokens": 10830970684,
-    "ultra_qa/blocks": 2430000,
-    "ultra_qa/targets": 3964462228,
-    "ultra_qa/input_tokens": 3964469848,
-    "cortex/blocks": 2430000,
-    "cortex/targets": 92442385,
-    "cortex/input_tokens": 3862348202,
-    "ultra_style/blocks": 2430000,
-    "ultra_style/targets": 3843618782,
-    "ultra_style/input_tokens": 3843820576,
-    "openmath/blocks": 1620000,
-    "openmath/targets": 2502933691,
-    "openmath/input_tokens": 2503767786,
-    "rewrite6/blocks": 810000,
-    "rewrite6/targets": 1327897186,
-    "rewrite6/input_tokens": 1329352974
+    "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
   },
   "sources": {
     "climbmix": {
-      "page": 888,
-      "row": 109,
+      "page": 912,
+      "row": 2733,
       "start": {
-        "file": 42,
+        "file": 44,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 75,
+          "row_group": 7,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 3637357
+      "offset": 6144,
+      "documents": 3738285
     },
     "rewrite6": {
-      "page": 4,
-      "row": 3530,
+      "page": 18,
+      "row": 4079,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 16384
+          "row": 73728
         }
       },
       "epoch": 10,
       "offset": 0,
-      "documents": 2073954
+      "documents": 2131847
     },
     "openmath": {
-      "page": 554,
-      "row": 2889,
+      "page": 570,
+      "row": 592,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 85,
-          "row": 934
+          "row_group": 151,
+          "row": 470
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2272073
+      "documents": 2335312
     },
     "ultra_qa": {
-      "page": 239,
-      "row": 3318,
+      "page": 246,
+      "row": 1981,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 7467
+          "row": 36139
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 982262
+      "documents": 1009597
     },
     "ultra_style": {
-      "page": 366,
-      "row": 262,
+      "page": 376,
+      "row": 1396,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 43735
+          "row": 84695
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1499398
+      "offset": 0,
+      "documents": 1541492
     },
     "cortex": {
-      "page": 47122,
-      "row": 22,
+      "page": 48431,
+      "row": 115,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 6031638
+      "documents": 6199283
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 180000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "68121455268768b4b43f0edce892dc1b160432a06c7d9a6bf2a59971a05dac46", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "25a0a70fc0e8c2815d5f59181512453873f858013da77228c792c3eefdaa0f65", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "fa0c68dff9ecb9880754b350e4c1810c4b7973ad3c5ba843eff2dd61df715747", "coverage.json": "8559bd7833f4eb51cd101d3f44a0ee6e30b948a85612ba19d9f1f4a93d4321de", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000180000"}
+{"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"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68121455268768b4b43f0edce892dc1b160432a06c7d9a6bf2a59971a05dac46
+oid sha256:6c683f5fcca897a05f7e0948afb2a85c2f5f99d3a2df8fe3ad69849852360942
 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": 75085.65875372152,
-      "required_pages_per_second_upper_estimate": 0.006508234597910304
+      "required_token_positions_per_second": 76375.43340313055,
+      "required_pages_per_second_upper_estimate": 0.006620028995617115
     },
     {
       "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": 9385.70734421519,
-      "required_pages_per_second_upper_estimate": 0.0037440989884375262
+      "required_token_positions_per_second": 9546.929175391318,
+      "required_pages_per_second_upper_estimate": 0.0038084127873748677
     },
     {
       "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": 18771.41468843038,
-      "required_pages_per_second_upper_estimate": 0.004342436422553239
+      "required_token_positions_per_second": 19093.858350782637,
+      "required_pages_per_second_upper_estimate": 0.004417028089023795
     },
     {
       "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": 28157.122032645573,
-      "required_pages_per_second_upper_estimate": 0.001714621002785909
+      "required_token_positions_per_second": 28640.787526173957,
+      "required_pages_per_second_upper_estimate": 0.0017440736937450589
     },
     {
       "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": 28157.122032645573,
-      "required_pages_per_second_upper_estimate": 0.002167613197975474
+      "required_token_positions_per_second": 28640.787526173957,
+      "required_pages_per_second_upper_estimate": 0.002204847106539067
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "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": 28157.122032645573,
-      "required_pages_per_second_upper_estimate": 0.34961721982002775
+      "required_token_positions_per_second": 28640.787526173957,
+      "required_pages_per_second_upper_estimate": 0.3556227265253729
     }
   ],
   "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": 62,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.08081432175779923,
-    "producer_p95_seconds": 0.34906973999750335,
-    "gpu_update_seconds": 0.981918534427798,
-    "producer_batches_per_second": 12.374044330867529,
-    "gpu_batches_per_second": 1.0184144253705718,
-    "producer_headroom": 12.150303474310046,
+    "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,
     "sustainable": true,
-    "buffer_seconds": 60.878949134523474,
+    "buffer_seconds": 60.81620480610735,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 506343.7655790157,
-    "upload_bytes_per_second": 5487182.841082292,
+    "generation_bytes_per_second": 283448.8065840012,
+    "upload_bytes_per_second": 4514531.554258477,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 108.64958501365523,
+    "estimated_full_shards_per_hour": 60.821515542412065,
     "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": 1.1056661020011234,
-      "bytes": 6106174,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0782650920009473,
-      "bytes": 5102757,
-      "pages": 36,
+      "seconds": 1.2709916069979954,
+      "bytes": 5895393,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0556685640003707,
-      "bytes": 7468948,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.2639917810010957,
+      "bytes": 13243059,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.1485331379990384,
-      "bytes": 5965663,
-      "pages": 28,
+      "seconds": 1.182673250001244,
+      "bytes": 5984869,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.951670551999996,
-      "bytes": 914174,
-      "pages": 26,
+      "seconds": 0.7554075879997981,
+      "bytes": 949451,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.4923641690002114,
-      "bytes": 17877863,
-      "pages": 40,
-      "shards": 5
+      "seconds": 1.1886153279992868,
+      "bytes": 11793828,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 0.7990716060012346,
-      "bytes": 1163166,
+      "seconds": 1.1566395750014635,
+      "bytes": 1136088,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.5562365400001,
-      "bytes": 7212570,
-      "pages": 29,
+      "seconds": 1.3760789849984576,
+      "bytes": 7774763,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9564802560016687,
-      "bytes": 6117579,
+      "seconds": 1.7174801740002295,
+      "bytes": 11887929,
       "pages": 34,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.0467319720009982,
-      "bytes": 5732511,
-      "pages": 41,
+      "seconds": 1.0596165809984086,
+      "bytes": 5348616,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.2224297300017497,
-      "bytes": 7561441,
-      "pages": 29,
+      "seconds": 1.3328053339973849,
+      "bytes": 7544008,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.1380056220004917,
-      "bytes": 6964732,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.952769712999725,
+      "bytes": 1151127,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.278426272001525,
-      "bytes": 5758474,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.7508555999993405,
+      "bytes": 8239169,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1922402710006281,
-      "bytes": 10988746,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.4154862130017136,
+      "bytes": 9920178,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 0.8872451259994705,
-      "bytes": 1248151,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.3325980400004482,
+      "bytes": 6856977,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.248066406998987,
-      "bytes": 7262597,
-      "pages": 37,
+      "seconds": 1.1581080529977044,
+      "bytes": 7348060,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 0.8794477230003395,
-      "bytes": 663465,
-      "pages": 19,
+      "seconds": 1.0286378179989697,
+      "bytes": 1051160,
+      "pages": 32,
       "shards": 1
     },
     {
-      "seconds": 1.393969371001731,
-      "bytes": 16331456,
-      "pages": 41,
-      "shards": 4
+      "seconds": 1.0477113340020878,
+      "bytes": 9989334,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.0225603920007416,
-      "bytes": 7736202,
-      "pages": 39,
+      "seconds": 1.1226200919991243,
+      "bytes": 7462663,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0140705839985458,
-      "bytes": 1147707,
-      "pages": 33,
+      "seconds": 0.8068668020023324,
+      "bytes": 982732,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.227832770000532,
-      "bytes": 5733304,
-      "pages": 21,
-      "shards": 2
+      "seconds": 1.4669019489992934,
+      "bytes": 13788589,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.3411155760004476,
-      "bytes": 6738025,
-      "pages": 38,
+      "seconds": 1.348856757998874,
+      "bytes": 5345336,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.1425984330016945,
-      "bytes": 6317215,
-      "pages": 35,
+      "seconds": 1.6173429520022182,
+      "bytes": 6229619,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.939828511000087,
-      "bytes": 6670718,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.705037297000672,
+      "bytes": 7666167,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.560656753998046,
-      "bytes": 11939208,
-      "pages": 42,
-      "shards": 4
+      "seconds": 0.7579026999992493,
+      "bytes": 928826,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.4417221109979437,
-      "bytes": 5916590,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.3881743410020135,
+      "bytes": 10608461,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 1.0029095059981046,
-      "bytes": 2598126,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.1401566910026304,
+      "bytes": 8280901,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.312767798997811,
-      "bytes": 6203437,
+      "seconds": 1.1491751960020338,
+      "bytes": 6275126,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.8579733459991985,
-      "bytes": 830240,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.3218376940021699,
+      "bytes": 6023489,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.2817783369973768,
-      "bytes": 11464455,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0548232209985144,
+      "bytes": 2410136,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.756591566001589,
-      "bytes": 11817435,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.1507087749996572,
+      "bytes": 10108802,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 0.9047177330030536,
-      "bytes": 2631324,
-      "pages": 36,
+      "seconds": 0.8633549010010029,
+      "bytes": 2675452,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3315005119984562,
-      "bytes": 5609510,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.1101803599995037,
+      "bytes": 11781038,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.1360651520008105,
-      "bytes": 10573263,
-      "pages": 40,
+      "seconds": 1.124476634999155,
+      "bytes": 9801537,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.0926981910015456,
-      "bytes": 2481076,
-      "pages": 31,
+      "seconds": 0.9195323169988114,
+      "bytes": 2651824,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3431768080008624,
-      "bytes": 5992564,
-      "pages": 32,
+      "seconds": 1.357327001998783,
+      "bytes": 5924306,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2225116730005539,
-      "bytes": 6709538,
-      "pages": 29,
+      "seconds": 0.9212070710018452,
+      "bytes": 2573619,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.5987381729974004,
-      "bytes": 11965797,
+      "seconds": 1.2234259870019741,
+      "bytes": 10476491,
       "pages": 39,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 1.2179241410012764,
-      "bytes": 6159934,
-      "pages": 31,
+      "seconds": 1.209228235999035,
+      "bytes": 11665620,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 0.8683807099987462,
+      "bytes": 2266346,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.8550500759993156,
-      "bytes": 2622896,
+      "seconds": 0.9792534570005955,
+      "bytes": 6085076,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1337759699999879,
-      "bytes": 5871486,
-      "pages": 32,
+      "seconds": 1.2787357830020483,
+      "bytes": 10476048,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3773436970004695,
+      "bytes": 2708180,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2709916069979954,
-      "bytes": 5895393,
-      "pages": 35,
+      "seconds": 1.157878524998523,
+      "bytes": 6108198,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2639917810010957,
-      "bytes": 13243059,
-      "pages": 34,
+      "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,
-      "shards": 3
-    },
-    {
-      "seconds": 1.6173429520022182,
-      "bytes": 6229619,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.705037297000672,
-      "bytes": 7666167,
-      "pages": 37,
-      "shards": 3
+      "seconds": 0.9504098300021724,
+      "bytes": 828896,
+      "pages": 27,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:25a0a70fc0e8c2815d5f59181512453873f858013da77228c792c3eefdaa0f65
-size 721776959
+oid sha256:267a2d850846babea151e8cb265757c9351aed7191d41054613399dd488e4d56
+size 721772031

training_config.json

@@ -46,7 +46,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 62,
+  "prefetch_batches": 63,
   "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,