Qwen90M step 110000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 105000,
+  "step": 110000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "2e7b636c49fbc9ff9f3020bb49601a599167b2c73e0d7cc4e1006453e0b1f4bb",
+    "model.safetensors": "771a846dfcaa9ac2239f4c06cbb98a3687803f55e50cb2bc31108644214bc872",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "7c54110ffbfd70520fd96fac4faa05d1b2f4692f233c1cb56199f384dddec7e5",
-    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
+    "trainer.pt": "847a0836b0889eaaab0298c4b432da9e790a6c0dff6648e498d3e21db336447a",
+    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
     "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98",
-    "shard_scheduler.json": "5bd2f77add256b6e8a0557483b02a1356ccca65bb941a3fd80d4f8230fce91c9",
-    "coverage.json": "e1d90370de539b285245087c6c1203bfa12344b8fb425a1eaa35cdbbf5795812",
+    "shard_scheduler.json": "c064c98867fe8d56cbdc4767e22ae652e692137319af0d6e6bccc2225e28dd1b",
+    "coverage.json": "2ca6a281fc319220e745c11a5bb944816bba2680cf6eeb45cee83e9353622a10",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 105000,
+  "step": 110000,
   "stats": {
-    "climbmix/blocks": 3780000,
-    "climbmix/targets": 6315746226,
-    "climbmix/input_tokens": 6316146929,
-    "ultra_qa/blocks": 1417500,
-    "ultra_qa/targets": 2312418151,
-    "ultra_qa/input_tokens": 2312422626,
-    "cortex/blocks": 1417500,
-    "cortex/targets": 53929281,
-    "cortex/input_tokens": 2253084355,
-    "ultra_style/blocks": 1417500,
-    "ultra_style/targets": 2242078536,
-    "ultra_style/input_tokens": 2242195826,
-    "openmath/blocks": 945000,
-    "openmath/targets": 1460177997,
-    "openmath/input_tokens": 1460664808,
-    "rewrite6/blocks": 472500,
-    "rewrite6/targets": 774617263,
-    "rewrite6/input_tokens": 775466653
+    "climbmix/blocks": 3960000,
+    "climbmix/targets": 6616550128,
+    "climbmix/input_tokens": 6616969711,
+    "ultra_qa/blocks": 1485000,
+    "ultra_qa/targets": 2422404971,
+    "ultra_qa/input_tokens": 2422409661,
+    "cortex/blocks": 1485000,
+    "cortex/targets": 56496990,
+    "cortex/input_tokens": 2360431026,
+    "ultra_style/blocks": 1485000,
+    "ultra_style/targets": 2348057065,
+    "ultra_style/input_tokens": 2348180586,
+    "openmath/blocks": 990000,
+    "openmath/targets": 1529649447,
+    "openmath/input_tokens": 1530159661,
+    "rewrite6/blocks": 495000,
+    "rewrite6/targets": 811489243,
+    "rewrite6/input_tokens": 812378858
   },
   "sources": {
     "climbmix": {
-      "page": 518,
-      "row": 1264,
+      "page": 542,
+      "row": 3742,
       "start": {
-        "file": 25,
+        "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 0,
+          "row_group": 13,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2122992
+      "offset": 0,
+      "documents": 2223774
     },
     "rewrite6": {
-      "page": 44,
-      "row": 2788,
+      "page": 8,
+      "row": 2154,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 180224
+          "row": 32768
         }
       },
-      "epoch": 5,
+      "epoch": 6,
       "offset": 0,
-      "documents": 1210032
+      "documents": 1267346
     },
     "openmath": {
-      "page": 323,
-      "row": 2616,
+      "page": 339,
+      "row": 371,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 13,
-          "row": 58
+          "row_group": 78,
+          "row": 594
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 1325624
+      "documents": 1388915
     },
     "ultra_qa": {
-      "page": 139,
-      "row": 3775,
+      "page": 146,
+      "row": 2430,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 0,
-          "row": 49717
+          "row_group": 1,
+          "row": 13902
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 573119
+      "offset": 2048,
+      "documents": 600446
     },
     "ultra_style": {
-      "page": 213,
-      "row": 1835,
+      "page": 224,
+      "row": 65,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 2,
-          "row": 16825
+          "row": 61881
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 874283
+      "documents": 917569
     },
     "cortex": {
-      "page": 27486,
-      "row": 47,
+      "page": 28796,
+      "row": 61,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 3518255
+      "documents": 3685949
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 105000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2e7b636c49fbc9ff9f3020bb49601a599167b2c73e0d7cc4e1006453e0b1f4bb", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "7c54110ffbfd70520fd96fac4faa05d1b2f4692f233c1cb56199f384dddec7e5", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98", "shard_scheduler.json": "5bd2f77add256b6e8a0557483b02a1356ccca65bb941a3fd80d4f8230fce91c9", "coverage.json": "e1d90370de539b285245087c6c1203bfa12344b8fb425a1eaa35cdbbf5795812", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000105000"}
+{"step": 110000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "771a846dfcaa9ac2239f4c06cbb98a3687803f55e50cb2bc31108644214bc872", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "847a0836b0889eaaab0298c4b432da9e790a6c0dff6648e498d3e21db336447a", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98", "shard_scheduler.json": "c064c98867fe8d56cbdc4767e22ae652e692137319af0d6e6bccc2225e28dd1b", "coverage.json": "2ca6a281fc319220e745c11a5bb944816bba2680cf6eeb45cee83e9353622a10", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000110000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2e7b636c49fbc9ff9f3020bb49601a599167b2c73e0d7cc4e1006453e0b1f4bb
+oid sha256:771a846dfcaa9ac2239f4c06cbb98a3687803f55e50cb2bc31108644214bc872
 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": 74828.03381926876,
-      "required_pages_per_second_upper_estimate": 0.006485904321536365
+      "required_token_positions_per_second": 75187.87582822153,
+      "required_pages_per_second_upper_estimate": 0.006517094514861157
     },
     {
       "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": 9353.504227408595,
-      "required_pages_per_second_upper_estimate": 0.0037312526836638725
+      "required_token_positions_per_second": 9398.484478527691,
+      "required_pages_per_second_upper_estimate": 0.0037491959783499643
     },
     {
       "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": 18707.00845481719,
-      "required_pages_per_second_upper_estimate": 0.004327537173917825
+      "required_token_positions_per_second": 18796.968957055382,
+      "required_pages_per_second_upper_estimate": 0.0043483479528598655
     },
     {
       "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": 28060.512682225788,
-      "required_pages_per_second_upper_estimate": 0.0017087380002154327
+      "required_token_positions_per_second": 28195.453435583073,
+      "required_pages_per_second_upper_estimate": 0.001716955184115474
     },
     {
       "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": 28060.512682225788,
-      "required_pages_per_second_upper_estimate": 0.002160175942748361
+      "required_token_positions_per_second": 28195.453435583073,
+      "required_pages_per_second_upper_estimate": 0.0021705640554817176
     },
     {
       "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": 28060.512682225788,
-      "required_pages_per_second_upper_estimate": 0.3484176550185106
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28195.453435583073,
+      "required_pages_per_second_upper_estimate": 0.3500931675575583
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 61,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.09439860396102517,
-    "producer_p95_seconds": 0.32911561600121786,
-    "gpu_update_seconds": 0.9852991751470356,
-    "producer_batches_per_second": 10.593376999652188,
-    "gpu_batches_per_second": 1.0149201635643008,
-    "producer_headroom": 10.437645619778879,
+    "producer_batch_seconds": 0.08555045376556336,
+    "producer_p95_seconds": 0.3201156809955137,
+    "gpu_update_seconds": 0.9805836271853611,
+    "producer_batches_per_second": 11.689008719232886,
+    "gpu_batches_per_second": 1.0198008331735775,
+    "producer_headroom": 11.462050568106696,
     "sustainable": true,
-    "buffer_seconds": 60.10324968396917,
+    "buffer_seconds": 60.79618488549239,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 85,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 488633.5744452333,
-    "upload_bytes_per_second": 4102653.020833448,
+    "generation_bytes_per_second": 390327.88351984415,
+    "upload_bytes_per_second": 4067954.1604864304,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 104.84939026849507,
+    "estimated_full_shards_per_hour": 83.75527743526929,
     "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": 1.0569672790006734,
-      "bytes": 5933537,
-      "pages": 27,
+      "seconds": 1.0622286870020616,
+      "bytes": 2500050,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.8821923329996935,
-      "bytes": 961115,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2227748900004372,
-      "bytes": 12637313,
-      "pages": 46,
-      "shards": 4
+      "seconds": 1.1363001400022767,
+      "bytes": 9937212,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.1576985259998764,
-      "bytes": 6035027,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.7940713860007236,
+      "bytes": 1067006,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.131166884999402,
-      "bytes": 11754261,
-      "pages": 32,
+      "seconds": 0.9383102729989332,
+      "bytes": 7322897,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.3580378289989312,
-      "bytes": 2581856,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1287910210012342,
+      "bytes": 10964395,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.6505394759988121,
-      "bytes": 9887447,
-      "pages": 28,
+      "seconds": 1.271768777998659,
+      "bytes": 8367852,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 0.7642709539977659,
-      "bytes": 2553358,
-      "pages": 36,
+      "seconds": 1.080213831999572,
+      "bytes": 6038268,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.2285513930000889,
-      "bytes": 6249359,
+      "seconds": 0.7658507999985886,
+      "bytes": 1224241,
       "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8377171489992179,
-      "bytes": 5823773,
-      "pages": 30,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.6930653950003034,
-      "bytes": 12802437,
-      "pages": 34,
+      "seconds": 34.26547425300305,
+      "bytes": 7641108,
+      "pages": 29,
       "shards": 4
     },
     {
-      "seconds": 1.2719050730011077,
-      "bytes": 5991225,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.169081484000344,
+      "bytes": 10839288,
+      "pages": 47,
+      "shards": 3
     },
     {
-      "seconds": 0.9874920929978543,
-      "bytes": 1116876,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.171269268001197,
+      "bytes": 2613392,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0614115569987916,
-      "bytes": 7614945,
-      "pages": 37,
+      "seconds": 1.150122408002062,
+      "bytes": 11429829,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.0503759480016015,
-      "bytes": 9758074,
-      "pages": 29,
+      "seconds": 1.3652685859997291,
+      "bytes": 10381816,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.035596928002633,
-      "bytes": 2562080,
+      "seconds": 1.0450683730014134,
+      "bytes": 2703655,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3296433830000751,
-      "bytes": 11732086,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.9104454030020861,
+      "bytes": 6024195,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.8983965400002489,
-      "bytes": 906815,
-      "pages": 27,
+      "seconds": 1.20260026500182,
+      "bytes": 1126978,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.218643701999099,
-      "bytes": 11866314,
-      "pages": 38,
-      "shards": 5
+      "seconds": 1.128464265002549,
+      "bytes": 12438546,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.0883352769997146,
-      "bytes": 6129558,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.0890558419996523,
+      "bytes": 11718926,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.048354777998611,
-      "bytes": 2655724,
+      "seconds": 0.9750485680015117,
+      "bytes": 2678604,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.8924994240005617,
-      "bytes": 5699054,
-      "pages": 19,
+      "seconds": 0.9471035349997692,
+      "bytes": 5896771,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.2686124620013288,
-      "bytes": 11642195,
-      "pages": 44,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9798610010002449,
-      "bytes": 7526317,
-      "pages": 37,
-      "shards": 3
+      "seconds": 0.970585907001805,
+      "bytes": 1087871,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 0.7459925380026107,
-      "bytes": 508402,
-      "pages": 17,
-      "shards": 1
+      "seconds": 1.300397458999214,
+      "bytes": 11506344,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 0.9750640929996734,
-      "bytes": 6281942,
+      "seconds": 1.0356865150024532,
+      "bytes": 6075433,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0568676230031997,
-      "bytes": 6659990,
-      "pages": 30,
-      "shards": 3
+      "seconds": 0.8513431389983452,
+      "bytes": 631745,
+      "pages": 18,
+      "shards": 1
     },
     {
-      "seconds": 1.1445812689999002,
-      "bytes": 6216290,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1167246349978086,
+      "bytes": 13461458,
+      "pages": 33,
+      "shards": 4
     },
     {
-      "seconds": 1.002836310002749,
-      "bytes": 6201464,
-      "pages": 35,
+      "seconds": 1.146249319997878,
+      "bytes": 4617912,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.0725969490013085,
-      "bytes": 8132406,
-      "pages": 29,
-      "shards": 3
+      "seconds": 0.8623779709996597,
+      "bytes": 6286843,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.011032721002266,
-      "bytes": 10718239,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.446577980001166,
+      "bytes": 2379048,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.0137605279996933,
-      "bytes": 6019285,
-      "pages": 27,
+      "seconds": 1.0237194600013027,
+      "bytes": 6170294,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 0.9841479640017496,
-      "bytes": 2601592,
+      "seconds": 1.2663461060001282,
+      "bytes": 11989707,
       "pages": 36,
+      "shards": 4
+    },
+    {
+      "seconds": 0.9998438940019696,
+      "bytes": 6675780,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1784665140003199,
-      "bytes": 6252756,
+      "seconds": 1.0791062770003919,
+      "bytes": 6176898,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9329704880001373,
-      "bytes": 2137944,
-      "pages": 23,
+      "seconds": 0.9380000849996577,
+      "bytes": 2661456,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.5199417449985049,
-      "bytes": 16335206,
-      "pages": 47,
-      "shards": 4
+      "seconds": 1.1787855490001675,
+      "bytes": 9821385,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 0.9530525390000548,
-      "bytes": 948458,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.015233548998367,
+      "bytes": 5854223,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.2729225670009328,
-      "bytes": 7447736,
+      "seconds": 0.8780484679991787,
+      "bytes": 2637899,
       "pages": 37,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.1847233869993943,
-      "bytes": 10716114,
+      "seconds": 1.1060435590006819,
+      "bytes": 11753109,
       "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.124874531000387,
-      "bytes": 2649444,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.0484200940009032,
+      "bytes": 6644677,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 0.9738625030004187,
-      "bytes": 6071559,
-      "pages": 35,
+      "seconds": 0.9731116549992294,
+      "bytes": 6198048,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5427607269994041,
-      "bytes": 11501779,
-      "pages": 29,
+      "seconds": 1.2161200219998136,
+      "bytes": 5847296,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.0622286870020616,
-      "bytes": 2500050,
-      "pages": 30,
+      "seconds": 1.0819334130028437,
+      "bytes": 2586941,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1363001400022767,
-      "bytes": 9937212,
+      "seconds": 1.1623649149987614,
+      "bytes": 11298342,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.7940713860007236,
-      "bytes": 1067006,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.995095141002821,
+      "bytes": 6689673,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 0.9383102729989332,
-      "bytes": 7322897,
-      "pages": 29,
+      "seconds": 0.9397628990009252,
+      "bytes": 7665294,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1287910210012342,
-      "bytes": 10964395,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.2407093820002046,
+      "bytes": 5771889,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.271768777998659,
-      "bytes": 8367852,
-      "pages": 34,
+      "seconds": 1.0350117630005116,
+      "bytes": 6519406,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.080213831999572,
-      "bytes": 6038268,
-      "pages": 31,
+      "seconds": 1.0306341840005189,
+      "bytes": 6113692,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.7658507999985886,
-      "bytes": 1224241,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.1684751539978606,
+      "bytes": 6202267,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 34.26547425300305,
-      "bytes": 7641108,
-      "pages": 29,
-      "shards": 4
+      "seconds": 1.093655648001004,
+      "bytes": 8128170,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.169081484000344,
-      "bytes": 10839288,
-      "pages": 47,
+      "seconds": 1.1689348799991421,
+      "bytes": 10022604,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.171269268001197,
-      "bytes": 2613392,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.8061036489998514,
+      "bytes": 878150,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.150122408002062,
-      "bytes": 11429829,
-      "pages": 30,
+      "seconds": 1.0866164300023229,
+      "bytes": 7550162,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.3652685859997291,
-      "bytes": 10381816,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.2529126609988452,
+      "bytes": 5669267,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.0450683730014134,
-      "bytes": 2703655,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.9512047030002577,
+      "bytes": 7497853,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 0.9104454030020861,
-      "bytes": 6024195,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1582007809993229,
+      "bytes": 12016223,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.20260026500182,
-      "bytes": 1126978,
-      "pages": 33,
+      "seconds": 0.9243304510018788,
+      "bytes": 1103405,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.128464265002549,
-      "bytes": 12438546,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.4641582209987973,
+      "bytes": 7254109,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.0890558419996523,
-      "bytes": 11718926,
-      "pages": 32,
+      "seconds": 1.1067068009979266,
+      "bytes": 10335265,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9750485680015117,
-      "bytes": 2678604,
-      "pages": 36,
+      "seconds": 1.2287768099959067,
+      "bytes": 2576063,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.9471035349997692,
-      "bytes": 5896771,
-      "pages": 27,
+      "seconds": 1.1285306259960635,
+      "bytes": 5981374,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.970585907001805,
-      "bytes": 1087871,
-      "pages": 35,
+      "seconds": 0.8470438740041573,
+      "bytes": 878239,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.300397458999214,
-      "bytes": 11506344,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.8046116949990392,
+      "bytes": 17147881,
+      "pages": 44,
+      "shards": 5
+    },
+    {
+      "seconds": 0.9106033309944905,
+      "bytes": 774779,
+      "pages": 26,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1979517419968033,
+      "bytes": 6268819,
+      "pages": 37,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c54110ffbfd70520fd96fac4faa05d1b2f4692f233c1cb56199f384dddec7e5
-size 721776575
+oid sha256:847a0836b0889eaaab0298c4b432da9e790a6c0dff6648e498d3e21db336447a
+size 721776063

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,