Qwen90M step 105000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 100000,
+  "step": 105000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "7e19e918403cc37f5d1810c36858d079fd9a1abc5258ce8bcadd60f8273dbc27",
+    "model.safetensors": "2e7b636c49fbc9ff9f3020bb49601a599167b2c73e0d7cc4e1006453e0b1f4bb",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "daa6ed4f3041a7768410d18fc4c1955d5f77761c3177da20014d301a73cb18e5",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "7c54110ffbfd70520fd96fac4faa05d1b2f4692f233c1cb56199f384dddec7e5",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98",
-    "shard_scheduler.json": "bedd197b1b3543492a228dfee98da9e5c2b64ac839bdf45da0854830f0f34e05",
-    "coverage.json": "807feed36aee1f42950d924959cfc4503b766399e37994c3a9760018c4ca6080",
+    "shard_scheduler.json": "5bd2f77add256b6e8a0557483b02a1356ccca65bb941a3fd80d4f8230fce91c9",
+    "coverage.json": "e1d90370de539b285245087c6c1203bfa12344b8fb425a1eaa35cdbbf5795812",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 100000,
+  "step": 105000,
   "stats": {
-    "climbmix/blocks": 3600000,
-    "climbmix/targets": 6015343871,
-    "climbmix/input_tokens": 6015725716,
-    "ultra_qa/blocks": 1350000,
-    "ultra_qa/targets": 2202272463,
-    "ultra_qa/input_tokens": 2202276743,
-    "cortex/blocks": 1350000,
-    "cortex/targets": 51355779,
-    "cortex/input_tokens": 2145762103,
-    "ultra_style/blocks": 1350000,
-    "ultra_style/targets": 2134287106,
-    "ultra_style/input_tokens": 2134399250,
-    "openmath/blocks": 900000,
-    "openmath/targets": 1390635021,
-    "openmath/input_tokens": 1391098627,
-    "rewrite6/blocks": 450000,
-    "rewrite6/targets": 737630929,
-    "rewrite6/input_tokens": 738439976
+    "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
   },
   "sources": {
     "climbmix": {
-      "page": 493,
-      "row": 2616,
+      "page": 518,
+      "row": 1264,
       "start": {
-        "file": 23,
+        "file": 25,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 66,
+          "row_group": 0,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2021944
+      "documents": 2122992
     },
     "rewrite6": {
-      "page": 30,
-      "row": 2677,
+      "page": 44,
+      "row": 2788,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 122880
+          "row": 180224
         }
       },
       "epoch": 5,
       "offset": 0,
-      "documents": 1152577
+      "documents": 1210032
     },
     "openmath": {
-      "page": 308,
-      "row": 982,
+      "page": 323,
+      "row": 2616,
       "start": {
-        "file": 2,
+        "file": 3,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 388,
-          "row": 268
+          "row_group": 13,
+          "row": 58
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 1262550
+      "documents": 1325624
     },
     "ultra_qa": {
-      "page": 133,
-      "row": 1106,
+      "page": 139,
+      "row": 3775,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 25141
+          "row": 49717
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 545874
+      "offset": 0,
+      "documents": 573119
     },
     "ultra_style": {
-      "page": 203,
-      "row": 2560,
+      "page": 213,
+      "row": 1835,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 61872
+          "row_group": 2,
+          "row": 16825
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 834048
+      "offset": 0,
+      "documents": 874283
     },
     "cortex": {
-      "page": 26176,
-      "row": 120,
+      "page": 27486,
+      "row": 47,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 3350648
+      "documents": 3518255
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 100000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "7e19e918403cc37f5d1810c36858d079fd9a1abc5258ce8bcadd60f8273dbc27", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "daa6ed4f3041a7768410d18fc4c1955d5f77761c3177da20014d301a73cb18e5", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98", "shard_scheduler.json": "bedd197b1b3543492a228dfee98da9e5c2b64ac839bdf45da0854830f0f34e05", "coverage.json": "807feed36aee1f42950d924959cfc4503b766399e37994c3a9760018c4ca6080", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000100000"}
+{"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"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7e19e918403cc37f5d1810c36858d079fd9a1abc5258ce8bcadd60f8273dbc27
+oid sha256:2e7b636c49fbc9ff9f3020bb49601a599167b2c73e0d7cc4e1006453e0b1f4bb
 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": 75313.11788406817,
-      "required_pages_per_second_upper_estimate": 0.006527950178838855
+      "required_token_positions_per_second": 74828.03381926876,
+      "required_pages_per_second_upper_estimate": 0.006485904321536365
     },
     {
       "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": 9414.139735508521,
-      "required_pages_per_second_upper_estimate": 0.003755441094426568
+      "required_token_positions_per_second": 9353.504227408595,
+      "required_pages_per_second_upper_estimate": 0.0037312526836638725
     },
     {
       "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": 18828.279471017042,
-      "required_pages_per_second_upper_estimate": 0.0043555910892186704
+      "required_token_positions_per_second": 18707.00845481719,
+      "required_pages_per_second_upper_estimate": 0.004327537173917825
     },
     {
       "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": 28242.419206525563,
-      "required_pages_per_second_upper_estimate": 0.0017198151531555152
+      "required_token_positions_per_second": 28060.512682225788,
+      "required_pages_per_second_upper_estimate": 0.0017087380002154327
     },
     {
       "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": 28242.419206525563,
-      "required_pages_per_second_upper_estimate": 0.0021741796105384454
+      "required_token_positions_per_second": 28060.512682225788,
+      "required_pages_per_second_upper_estimate": 0.002160175942748361
     },
     {
       "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": 28242.419206525563,
-      "required_pages_per_second_upper_estimate": 0.35067632524834
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 28060.512682225788,
+      "required_pages_per_second_upper_estimate": 0.3484176550185106
     }
   ],
   "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": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.11091515512487149,
-    "producer_p95_seconds": 0.30749590000050375,
-    "gpu_update_seconds": 0.978952964256397,
-    "producer_batches_per_second": 9.01590047702833,
-    "gpu_batches_per_second": 1.02149953727306,
-    "producer_headroom": 8.826142497427547,
+    "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,
     "sustainable": true,
-    "buffer_seconds": 60.695083783896614,
+    "buffer_seconds": 60.10324968396917,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 359571.1211645759,
-    "upload_bytes_per_second": 6007747.908546246,
+    "generation_bytes_per_second": 488633.5744452333,
+    "upload_bytes_per_second": 4102653.020833448,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 77.15559221461257,
+    "estimated_full_shards_per_hour": 104.84939026849507,
     "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": 1.214833362000718,
-      "bytes": 7500577,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9954615130009188,
-      "bytes": 6295663,
-      "pages": 36,
+      "seconds": 1.0569672790006734,
+      "bytes": 5933537,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0289003750003758,
-      "bytes": 6769832,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.8821923329996935,
+      "bytes": 961115,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.118036003999805,
-      "bytes": 11128347,
-      "pages": 36,
+      "seconds": 1.2227748900004372,
+      "bytes": 12637313,
+      "pages": 46,
       "shards": 4
     },
     {
-      "seconds": 1.0285996809998323,
-      "bytes": 1163689,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0598135079999338,
-      "bytes": 6175962,
+      "seconds": 1.1576985259998764,
+      "bytes": 6035027,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.1872953079982835,
-      "bytes": 7019215,
-      "pages": 28,
+      "seconds": 1.131166884999402,
+      "bytes": 11754261,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.280990704002761,
-      "bytes": 5669008,
-      "pages": 37,
+      "seconds": 1.3580378289989312,
+      "bytes": 2581856,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2822542080029962,
-      "bytes": 13299747,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.6505394759988121,
+      "bytes": 9887447,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 0.8506751649983926,
-      "bytes": 1134712,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.7642709539977659,
+      "bytes": 2553358,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1013248220006062,
-      "bytes": 6047733,
+      "seconds": 1.2285513930000889,
+      "bytes": 6249359,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2326910120027605,
-      "bytes": 11762073,
-      "pages": 32,
-      "shards": 4
+      "seconds": 0.8377171489992179,
+      "bytes": 5823773,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 0.9280554619981558,
-      "bytes": 1182136,
+      "seconds": 1.6930653950003034,
+      "bytes": 12802437,
       "pages": 34,
-      "shards": 1
+      "shards": 4
     },
     {
-      "seconds": 1.007938288999867,
-      "bytes": 6307262,
-      "pages": 35,
+      "seconds": 1.2719050730011077,
+      "bytes": 5991225,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1597189560015977,
-      "bytes": 13015122,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.9874920929978543,
+      "bytes": 1116876,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.0076978060023976,
-      "bytes": 5410849,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.0614115569987916,
+      "bytes": 7614945,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.4046865109994542,
-      "bytes": 7189609,
+      "seconds": 1.0503759480016015,
+      "bytes": 9758074,
       "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 0.8812644179997733,
-      "bytes": 1118965,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.006351252002787,
-      "bytes": 5851026,
-      "pages": 35,
+      "seconds": 1.035596928002633,
+      "bytes": 2562080,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3481963590020314,
-      "bytes": 12258822,
+      "seconds": 1.3296433830000751,
+      "bytes": 11732086,
       "pages": 34,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 0.9640109699976165,
-      "bytes": 6845574,
-      "pages": 31,
-      "shards": 2
+      "seconds": 0.8983965400002489,
+      "bytes": 906815,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.094309231000807,
-      "bytes": 5916543,
-      "pages": 36,
+      "seconds": 1.218643701999099,
+      "bytes": 11866314,
+      "pages": 38,
+      "shards": 5
+    },
+    {
+      "seconds": 1.0883352769997146,
+      "bytes": 6129558,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0784175260014308,
-      "bytes": 7549095,
+      "seconds": 1.048354777998611,
+      "bytes": 2655724,
       "pages": 36,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.0992059410018555,
-      "bytes": 4757047,
-      "pages": 31,
+      "seconds": 0.8924994240005617,
+      "bytes": 5699054,
+      "pages": 19,
       "shards": 2
     },
     {
-      "seconds": 0.9332335270009935,
-      "bytes": 581892,
-      "pages": 17,
-      "shards": 1
+      "seconds": 1.2686124620013288,
+      "bytes": 11642195,
+      "pages": 44,
+      "shards": 3
     },
     {
-      "seconds": 1.1391355480009224,
-      "bytes": 7817543,
+      "seconds": 0.9798610010002449,
+      "bytes": 7526317,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.0901677820002078,
-      "bytes": 6115749,
-      "pages": 34,
-      "shards": 2
+      "seconds": 0.7459925380026107,
+      "bytes": 508402,
+      "pages": 17,
+      "shards": 1
     },
     {
-      "seconds": 1.5552335219981615,
-      "bytes": 12252293,
-      "pages": 37,
-      "shards": 4
+      "seconds": 0.9750640929996734,
+      "bytes": 6281942,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 0.9495614830011618,
-      "bytes": 5780919,
-      "pages": 27,
+      "seconds": 1.0568676230031997,
+      "bytes": 6659990,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.0740993799990974,
-      "bytes": 5965207,
-      "pages": 35,
+      "seconds": 1.1445812689999002,
+      "bytes": 6216290,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9706554090007558,
-      "bytes": 2646898,
+      "seconds": 1.002836310002749,
+      "bytes": 6201464,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0982734540011734,
-      "bytes": 10287883,
-      "pages": 34,
+      "seconds": 1.0725969490013085,
+      "bytes": 8132406,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 0.8711374440026702,
-      "bytes": 2523849,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5673655419996066,
-      "bytes": 11589450,
-      "pages": 30,
+      "seconds": 1.011032721002266,
+      "bytes": 10718239,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.247284810997371,
-      "bytes": 6017710,
-      "pages": 36,
+      "seconds": 1.0137605279996933,
+      "bytes": 6019285,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 0.9974236470006872,
-      "bytes": 6556698,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.9841479640017496,
+      "bytes": 2601592,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0323645409989695,
-      "bytes": 5955575,
+      "seconds": 1.1784665140003199,
+      "bytes": 6252756,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.8832039090011676,
-      "bytes": 6189230,
-      "pages": 36,
+      "seconds": 0.9329704880001373,
+      "bytes": 2137944,
+      "pages": 23,
       "shards": 2
     },
     {
-      "seconds": 0.9038558969987207,
-      "bytes": 2334921,
-      "pages": 27,
-      "shards": 2
-    },
-    {
-      "seconds": 1.410665313997015,
-      "bytes": 15644643,
-      "pages": 39,
+      "seconds": 1.5199417449985049,
+      "bytes": 16335206,
+      "pages": 47,
       "shards": 4
     },
     {
-      "seconds": 0.8785766789987974,
-      "bytes": 2610651,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.9530525390000548,
+      "bytes": 948458,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.0569672790006734,
-      "bytes": 5933537,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.2729225670009328,
+      "bytes": 7447736,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 0.8821923329996935,
-      "bytes": 961115,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.1847233869993943,
+      "bytes": 10716114,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.2227748900004372,
-      "bytes": 12637313,
-      "pages": 46,
-      "shards": 4
+      "seconds": 1.124874531000387,
+      "bytes": 2649444,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1576985259998764,
-      "bytes": 6035027,
+      "seconds": 0.9738625030004187,
+      "bytes": 6071559,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.131166884999402,
-      "bytes": 11754261,
-      "pages": 32,
+      "seconds": 1.5427607269994041,
+      "bytes": 11501779,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.3580378289989312,
-      "bytes": 2581856,
-      "pages": 36,
+      "seconds": 1.0622286870020616,
+      "bytes": 2500050,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.6505394759988121,
-      "bytes": 9887447,
-      "pages": 28,
+      "seconds": 1.1363001400022767,
+      "bytes": 9937212,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.7642709539977659,
-      "bytes": 2553358,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.7940713860007236,
+      "bytes": 1067006,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.2285513930000889,
-      "bytes": 6249359,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.9383102729989332,
+      "bytes": 7322897,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 0.8377171489992179,
-      "bytes": 5823773,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.1287910210012342,
+      "bytes": 10964395,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.6930653950003034,
-      "bytes": 12802437,
+      "seconds": 1.271768777998659,
+      "bytes": 8367852,
       "pages": 34,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 1.2719050730011077,
-      "bytes": 5991225,
-      "pages": 33,
+      "seconds": 1.080213831999572,
+      "bytes": 6038268,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 0.9874920929978543,
-      "bytes": 1116876,
-      "pages": 34,
+      "seconds": 0.7658507999985886,
+      "bytes": 1224241,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.0614115569987916,
-      "bytes": 7614945,
-      "pages": 37,
-      "shards": 3
+      "seconds": 34.26547425300305,
+      "bytes": 7641108,
+      "pages": 29,
+      "shards": 4
     },
     {
-      "seconds": 1.0503759480016015,
-      "bytes": 9758074,
-      "pages": 29,
+      "seconds": 1.169081484000344,
+      "bytes": 10839288,
+      "pages": 47,
       "shards": 3
     },
     {
-      "seconds": 1.035596928002633,
-      "bytes": 2562080,
+      "seconds": 1.171269268001197,
+      "bytes": 2613392,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3296433830000751,
-      "bytes": 11732086,
-      "pages": 34,
+      "seconds": 1.150122408002062,
+      "bytes": 11429829,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.8983965400002489,
-      "bytes": 906815,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.3652685859997291,
+      "bytes": 10381816,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.218643701999099,
-      "bytes": 11866314,
-      "pages": 38,
-      "shards": 5
+      "seconds": 1.0450683730014134,
+      "bytes": 2703655,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0883352769997146,
-      "bytes": 6129558,
-      "pages": 35,
+      "seconds": 0.9104454030020861,
+      "bytes": 6024195,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.048354777998611,
-      "bytes": 2655724,
+      "seconds": 1.20260026500182,
+      "bytes": 1126978,
+      "pages": 33,
+      "shards": 1
+    },
+    {
+      "seconds": 1.128464265002549,
+      "bytes": 12438546,
+      "pages": 34,
+      "shards": 4
+    },
+    {
+      "seconds": 1.0890558419996523,
+      "bytes": 11718926,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "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": 0.970585907001805,
+      "bytes": 1087871,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.300397458999214,
+      "bytes": 11506344,
+      "pages": 35,
+      "shards": 4
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:daa6ed4f3041a7768410d18fc4c1955d5f77761c3177da20014d301a73cb18e5
-size 721768255
+oid sha256:7c54110ffbfd70520fd96fac4faa05d1b2f4692f233c1cb56199f384dddec7e5
+size 721776575

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 62,
-  "prefetch_low_water": 31,
+  "prefetch_batches": 61,
+  "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,