Qwen90M step 330000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 325000,
+  "step": 330000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde",
+    "model.safetensors": "328bd745ba3eedb0c09045707f1c3645741429105fa758b46400ed4e42af0708",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64",
-    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
+    "trainer.pt": "15e5868d1d2d9bba0011346b8b8fe6a7017e790885cbae0ddebf73fbc62c7dfe",
+    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
     "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a",
-    "shard_scheduler.json": "fcc8218f92e7a2b7ab38bc12e80b76d8e5163a9d89f4285f1044c2f3f6d27145",
-    "coverage.json": "877bcc297c777298fe977b7e68fa8d3e9711497b231150a19f8f8e0f118f5cab",
+    "shard_scheduler.json": "9ea0c54003a1a89d00999afcc33c19cd9e20fd18d66970efa7421ab898bf1d77",
+    "coverage.json": "8db65ea8720123ccbfed7d388c2266922b6f26d48d6d3b735a4714722ac5561c",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 325000,
+  "step": 330000,
   "stats": {
-    "climbmix/blocks": 11700000,
-    "climbmix/targets": 19557876823,
-    "climbmix/input_tokens": 19559116697,
-    "ultra_qa/blocks": 4387500,
-    "ultra_qa/targets": 7157752008,
-    "ultra_qa/input_tokens": 7157765811,
-    "cortex/blocks": 4387500,
-    "cortex/targets": 166932432,
-    "cortex/input_tokens": 6974752559,
-    "ultra_style/blocks": 4387500,
-    "ultra_style/targets": 6940030981,
-    "ultra_style/input_tokens": 6940394926,
-    "openmath/blocks": 2925000,
-    "openmath/targets": 4519854927,
-    "openmath/input_tokens": 4521361389,
-    "rewrite6/blocks": 1462500,
-    "rewrite6/targets": 2397565289,
-    "rewrite6/input_tokens": 2400193805
+    "climbmix/blocks": 11880000,
+    "climbmix/targets": 19858829729,
+    "climbmix/input_tokens": 19860088763,
+    "ultra_qa/blocks": 4455000,
+    "ultra_qa/targets": 7267913693,
+    "ultra_qa/input_tokens": 7267927725,
+    "cortex/blocks": 4455000,
+    "cortex/targets": 169500703,
+    "cortex/input_tokens": 7082136293,
+    "ultra_style/blocks": 4455000,
+    "ultra_style/targets": 7046969672,
+    "ultra_style/input_tokens": 7047339282,
+    "openmath/blocks": 2970000,
+    "openmath/targets": 4589449827,
+    "openmath/input_tokens": 4590979616,
+    "rewrite6/blocks": 1485000,
+    "rewrite6/targets": 2434435841,
+    "rewrite6/input_tokens": 2437105213
   },
   "sources": {
     "climbmix": {
-      "page": 1602,
-      "row": 3042,
+      "page": 1627,
+      "row": 1818,
       "start": {
-        "file": 77,
+        "file": 78,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 44,
+          "row_group": 62,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 6564834
+      "documents": 6666010
     },
     "rewrite6": {
-      "page": 11,
-      "row": 2354,
+      "page": 25,
+      "row": 3161,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 45056
+          "row": 102400
         }
       },
       "epoch": 18,
-      "offset": 2048,
-      "documents": 3744682
+      "offset": 0,
+      "documents": 3802833
     },
     "openmath": {
-      "page": 1001,
-      "row": 2216,
+      "page": 1016,
+      "row": 4063,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 170,
-          "row": 246
+          "row_group": 231,
+          "row": 686
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 4102312
+      "documents": 4165599
     },
     "ultra_qa": {
-      "page": 432,
-      "row": 4084,
+      "page": 439,
+      "row": 2714,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 3,
-          "row": 16720
+          "row": 45392
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1773556
+      "offset": 4096,
+      "documents": 1800858
     },
     "ultra_style": {
-      "page": 660,
-      "row": 3969,
+      "page": 671,
+      "row": 421,
       "start": {
-        "file": 3,
+        "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 51868
+          "row_group": 0,
+          "row": 8645
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 2707329
+      "offset": 2048,
+      "documents": 2748837
     },
     "cortex": {
-      "page": 85096,
-      "row": 43,
+      "page": 86407,
+      "row": 78,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 10892331
+      "documents": 11060174
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 325000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "fcc8218f92e7a2b7ab38bc12e80b76d8e5163a9d89f4285f1044c2f3f6d27145", "coverage.json": "877bcc297c777298fe977b7e68fa8d3e9711497b231150a19f8f8e0f118f5cab", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000325000"}
+{"step": 330000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "328bd745ba3eedb0c09045707f1c3645741429105fa758b46400ed4e42af0708", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "15e5868d1d2d9bba0011346b8b8fe6a7017e790885cbae0ddebf73fbc62c7dfe", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "9ea0c54003a1a89d00999afcc33c19cd9e20fd18d66970efa7421ab898bf1d77", "coverage.json": "8db65ea8720123ccbfed7d388c2266922b6f26d48d6d3b735a4714722ac5561c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000330000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde
+oid sha256:328bd745ba3eedb0c09045707f1c3645741429105fa758b46400ed4e42af0708
 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": 74569.75015202582,
-      "required_pages_per_second_upper_estimate": 0.006463516947873717
+      "required_token_positions_per_second": 75630.60254063262,
+      "required_pages_per_second_upper_estimate": 0.006555468944212357
     },
     {
       "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": 9321.218769003228,
-      "required_pages_per_second_upper_estimate": 0.0037183735315953516
+      "required_token_positions_per_second": 9453.825317579078,
+      "required_pages_per_second_upper_estimate": 0.003771272266466841
     },
     {
       "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": 18642.437538006456,
-      "required_pages_per_second_upper_estimate": 0.004312599828722984
+      "required_token_positions_per_second": 18907.650635158156,
+      "required_pages_per_second_upper_estimate": 0.004373952211157938
     },
     {
       "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": 27963.65630700968,
-      "required_pages_per_second_upper_estimate": 0.0017028399551308995
+      "required_token_positions_per_second": 28361.475952737233,
+      "required_pages_per_second_upper_estimate": 0.0017270650843573335
     },
     {
       "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": 27963.65630700968,
-      "required_pages_per_second_upper_estimate": 0.0021527196708686185
+      "required_token_positions_per_second": 28361.475952737233,
+      "required_pages_per_second_upper_estimate": 0.002183344928431987
     },
     {
       "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": 27963.65630700968,
-      "required_pages_per_second_upper_estimate": 0.34721502299576196
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28361.475952737233,
+      "required_pages_per_second_upper_estimate": 0.3521546115789914
     }
   ],
   "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": 690670,
-    "producer_batch_seconds": 0.074832562304735,
-    "producer_p95_seconds": 0.19214319799903024,
-    "gpu_update_seconds": 0.988711908645131,
-    "producer_batches_per_second": 13.363166637643321,
-    "gpu_batches_per_second": 1.011416967122746,
-    "producer_headroom": 13.212321991847265,
+    "producer_batch_seconds": 0.0792378665312583,
+    "producer_p95_seconds": 0.2171078139999736,
+    "gpu_update_seconds": 0.9748434829722473,
+    "producer_batches_per_second": 12.620228733764721,
+    "gpu_batches_per_second": 1.0258056985220354,
+    "producer_headroom": 12.302747734729635,
     "sustainable": true,
-    "buffer_seconds": 60.31142642735299,
+    "buffer_seconds": 60.440295944279335,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 299588.0660923766,
-    "upload_bytes_per_second": 4058653.93467418,
+    "generation_bytes_per_second": 523269.04561499885,
+    "upload_bytes_per_second": 4017522.9884006456,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 64.28462493017648,
+    "estimated_full_shards_per_hour": 112.28135610902285,
     "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": 1.8453115359998264,
-      "bytes": 9879354,
-      "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0517817119998654,
-      "bytes": 7714712,
+      "seconds": 1.6009747379994224,
+      "bytes": 7396060,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9509194960000968,
-      "bytes": 6981450,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.9060076790001403,
+      "bytes": 1026778,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.187183233000269,
-      "bytes": 9995831,
-      "pages": 33,
+      "seconds": 1.30779888699999,
+      "bytes": 10925777,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.3044787690000703,
-      "bytes": 7311631,
+      "seconds": 0.9117075719987042,
+      "bytes": 2359249,
       "pages": 28,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 0.7353289520001454,
-      "bytes": 1217005,
+      "seconds": 0.9851995050012192,
+      "bytes": 5930447,
       "pages": 34,
-      "shards": 1
+      "shards": 2
     },
     {
-      "seconds": 1.1706305770003382,
-      "bytes": 7901550,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.2710061699999642,
+      "bytes": 13577169,
+      "pages": 42,
+      "shards": 4
     },
     {
-      "seconds": 1.067005848000008,
-      "bytes": 5080800,
-      "pages": 32,
-      "shards": 2
+      "seconds": 0.8450371549988631,
+      "bytes": 895470,
+      "pages": 25,
+      "shards": 1
     },
     {
-      "seconds": 0.9038640889998533,
-      "bytes": 6072537,
-      "pages": 27,
-      "shards": 2
+      "seconds": 32.774606008000774,
+      "bytes": 10411787,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.318457908000255,
-      "bytes": 13323724,
-      "pages": 39,
-      "shards": 4
+      "seconds": 1.2439632269997674,
+      "bytes": 7901436,
+      "pages": 46,
+      "shards": 3
     },
     {
-      "seconds": 0.8108393100001194,
-      "bytes": 1203088,
+      "seconds": 0.8556265790011821,
+      "bytes": 1138175,
       "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.1118799440000657,
-      "bytes": 10497270,
+      "seconds": 1.1632403840012557,
+      "bytes": 11705717,
       "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.8437159310005882,
-      "bytes": 2537590,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.197896118999779,
+      "bytes": 12261236,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.248541894000482,
-      "bytes": 6395911,
-      "pages": 35,
+      "seconds": 0.8173304430001735,
+      "bytes": 1136030,
+      "pages": 33,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1651587139986077,
+      "bytes": 5801842,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.171668709000187,
-      "bytes": 7290583,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.0791701340003783,
+      "bytes": 2256184,
+      "pages": 26,
+      "shards": 2
     },
     {
-      "seconds": 1.2945431420002933,
-      "bytes": 10577919,
-      "pages": 37,
+      "seconds": 1.5194503539987636,
+      "bytes": 9946744,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2328179739997722,
-      "bytes": 5713232,
-      "pages": 27,
+      "seconds": 1.188283532999776,
+      "bytes": 2401103,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 0.9817969460000313,
-      "bytes": 2633705,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.182281362998765,
+      "bytes": 11896636,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.1478959790001682,
-      "bytes": 6105703,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.8353424750002887,
+      "bytes": 975468,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.3559997279999152,
-      "bytes": 10442134,
-      "pages": 30,
+      "seconds": 1.1824829600009252,
+      "bytes": 12495213,
+      "pages": 37,
+      "shards": 4
+    },
+    {
+      "seconds": 1.254413848000695,
+      "bytes": 6063292,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.8416516239994962,
-      "bytes": 2569575,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.9384880159996101,
+      "bytes": 1115458,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.165861663999749,
-      "bytes": 11609605,
-      "pages": 32,
+      "seconds": 1.1859958829991228,
+      "bytes": 7375397,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.5608764529997643,
-      "bytes": 2529116,
-      "pages": 35,
+      "seconds": 0.9071520359993883,
+      "bytes": 6539230,
+      "pages": 25,
       "shards": 2
     },
     {
-      "seconds": 1.0529914669996288,
-      "bytes": 5481895,
-      "pages": 19,
-      "shards": 2
+      "seconds": 1.2508507479997206,
+      "bytes": 11653692,
+      "pages": 45,
+      "shards": 4
     },
     {
-      "seconds": 1.0121831689993996,
-      "bytes": 5661667,
-      "pages": 37,
+      "seconds": 1.2195578610007942,
+      "bytes": 5944204,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0493596160004017,
-      "bytes": 7456331,
-      "pages": 30,
-      "shards": 3
+      "seconds": 0.9418964010001218,
+      "bytes": 1018742,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 0.9663659530006043,
-      "bytes": 5966016,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.3071213839994016,
+      "bytes": 12190710,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 0.7928353870001956,
-      "bytes": 647903,
-      "pages": 20,
+      "seconds": 0.9488432640009705,
+      "bytes": 921326,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.4357455670005947,
-      "bytes": 17610080,
+      "seconds": 1.0550517599986051,
+      "bytes": 5962735,
+      "pages": 26,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1898453599987988,
+      "bytes": 13419202,
       "pages": 43,
-      "shards": 5
+      "shards": 4
     },
     {
-      "seconds": 0.884776743000657,
-      "bytes": 1180635,
-      "pages": 33,
+      "seconds": 0.9201347599992005,
+      "bytes": 838883,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.2576917689993934,
-      "bytes": 6183813,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.1256958809990465,
+      "bytes": 12103363,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.3994414539993159,
-      "bytes": 7586010,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.0281697059999715,
+      "bytes": 1191797,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.0743332829997598,
-      "bytes": 4951926,
-      "pages": 31,
+      "seconds": 0.9877725879996433,
+      "bytes": 6235947,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3130787809996036,
-      "bytes": 11524150,
-      "pages": 30,
+      "seconds": 1.2409586350004247,
+      "bytes": 7163170,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.0262837310001487,
-      "bytes": 2640097,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.270394237999426,
+      "bytes": 10514630,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 1.0240621299999475,
-      "bytes": 6061414,
-      "pages": 35,
+      "seconds": 1.509260009999707,
+      "bytes": 6276691,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.345420446999924,
-      "bytes": 12056021,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.1396728859999712,
+      "bytes": 2416731,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.0655096850005066,
-      "bytes": 1134459,
+      "seconds": 0.9309691420003219,
+      "bytes": 6123756,
       "pages": 33,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1734728310002538,
-      "bytes": 6054797,
-      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2160326559996975,
-      "bytes": 2369678,
-      "pages": 28,
+      "seconds": 1.1473543519987288,
+      "bytes": 5736316,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.405440059999819,
-      "bytes": 15806866,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.0078334160007216,
+      "bytes": 7473663,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.6009747379994224,
-      "bytes": 7396060,
+      "seconds": 1.0918037890005508,
+      "bytes": 11651730,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9060076790001403,
-      "bytes": 1026778,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.0898228620008013,
+      "bytes": 2320314,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.30779888699999,
-      "bytes": 10925777,
-      "pages": 35,
+      "seconds": 0.9809586890005448,
+      "bytes": 9894024,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.9117075719987042,
-      "bytes": 2359249,
-      "pages": 28,
+      "seconds": 1.0132985079999344,
+      "bytes": 6275693,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9851995050012192,
-      "bytes": 5930447,
-      "pages": 34,
+      "seconds": 0.9494441940005345,
+      "bytes": 2648505,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2710061699999642,
-      "bytes": 13577169,
-      "pages": 42,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8450371549988631,
-      "bytes": 895470,
-      "pages": 25,
-      "shards": 1
+      "seconds": 1.0081078959992738,
+      "bytes": 5874669,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 32.774606008000774,
-      "bytes": 10411787,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.0266657669999404,
+      "bytes": 4655742,
+      "pages": 12,
+      "shards": 2
     },
     {
-      "seconds": 1.2439632269997674,
-      "bytes": 7901436,
-      "pages": 46,
-      "shards": 3
+      "seconds": 1.4716465549990971,
+      "bytes": 13803834,
+      "pages": 48,
+      "shards": 4
     },
     {
-      "seconds": 0.8556265790011821,
-      "bytes": 1138175,
-      "pages": 33,
-      "shards": 1
+      "seconds": 0.940834745999382,
+      "bytes": 6400703,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.1632403840012557,
-      "bytes": 11705717,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.1989260670015938,
+      "bytes": 2439937,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.197896118999779,
-      "bytes": 12261236,
-      "pages": 40,
+      "seconds": 1.2518158220009354,
+      "bytes": 10474054,
+      "pages": 38,
       "shards": 4
     },
     {
-      "seconds": 0.8173304430001735,
-      "bytes": 1136030,
-      "pages": 33,
+      "seconds": 1.1628889589992468,
+      "bytes": 1077101,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.1651587139986077,
-      "bytes": 5801842,
-      "pages": 20,
-      "shards": 2
+      "seconds": 1.117781933000515,
+      "bytes": 7831651,
+      "pages": 22,
+      "shards": 3
     },
     {
-      "seconds": 1.0791701340003783,
-      "bytes": 2256184,
-      "pages": 26,
+      "seconds": 1.0544361239990394,
+      "bytes": 7130727,
+      "pages": 44,
       "shards": 2
     },
     {
-      "seconds": 1.5194503539987636,
-      "bytes": 9946744,
+      "seconds": 1.1150904570004059,
+      "bytes": 10599538,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0015506020008615,
+      "bytes": 7818449,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.188283532999776,
-      "bytes": 2401103,
-      "pages": 29,
-      "shards": 2
+      "seconds": 0.8705007850003312,
+      "bytes": 1070987,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.182281362998765,
-      "bytes": 11896636,
-      "pages": 38,
+      "seconds": 0.9655942159988626,
+      "bytes": 7298211,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 0.8353424750002887,
-      "bytes": 975468,
-      "pages": 29,
+      "seconds": 0.8577407669999957,
+      "bytes": 1078743,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.1824829600009252,
-      "bytes": 12495213,
-      "pages": 37,
+      "seconds": 1.3106396379989746,
+      "bytes": 15458880,
+      "pages": 39,
       "shards": 4
     },
     {
-      "seconds": 1.254413848000695,
-      "bytes": 6063292,
-      "pages": 35,
+      "seconds": 0.9607973950005544,
+      "bytes": 7512102,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9384880159996101,
-      "bytes": 1115458,
-      "pages": 34,
+      "seconds": 2.3653303009996307,
+      "bytes": 1160564,
+      "pages": 33,
       "shards": 1
-    },
-    {
-      "seconds": 1.1859958829991228,
-      "bytes": 7375397,
-      "pages": 29,
-      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64
-size 721767807
+oid sha256:15e5868d1d2d9bba0011346b8b8fe6a7017e790885cbae0ddebf73fbc62c7dfe
+size 721778495

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,