Qwen90M step 305000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 300000,
+  "step": 305000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "6492e2662f94faa6a2e8ee1d628fd698c35d02f34d6f8f370f9d4db12434c3dc",
+    "model.safetensors": "89b7e1e53c78bd469c045acaa45bf2613b245ad7c172ee689268c146802eaa57",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "fbe139ff1c427ab9b5843d95ec3ecec658133f721c1f794977096107acee4acb",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "124b35ad649617fdf90d8ce9db8b59458f1bc588be0bbe9ced8a83035f7877a9",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "f89f82c3f4c44a52fcd532346ea71580902d757e59e1961e9b2bed0030259dba",
-    "shard_scheduler.json": "a2c401bec97cd3feb72edade9a3ae615ffe06fc28d4766333eb54ffc7af1ddd7",
-    "coverage.json": "251945ddb781b0d2536a5c2056932a9ec606d4b451ee97c2b33e28f6ace29956",
+    "shard_scheduler.json": "9022b90c7229c4a1b45c4f0abd2854b2ee887f5b962134d4b9e98bb6007410cc",
+    "coverage.json": "156ed6483cb0396d79e81df2dfaa6ef016a16020c5c9fcc2a445e0bcff68144f",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 300000,
+  "step": 305000,
   "stats": {
-    "climbmix/blocks": 10800000,
-    "climbmix/targets": 18053965003,
-    "climbmix/input_tokens": 18055109204,
-    "ultra_qa/blocks": 4050000,
-    "ultra_qa/targets": 6607020438,
-    "ultra_qa/input_tokens": 6607033152,
-    "cortex/blocks": 4050000,
-    "cortex/targets": 154079984,
-    "cortex/input_tokens": 6438099093,
-    "ultra_style/blocks": 4050000,
-    "ultra_style/targets": 6406628156,
-    "ultra_style/input_tokens": 6406964117,
-    "openmath/blocks": 2700000,
-    "openmath/targets": 4172214215,
-    "openmath/input_tokens": 4173605124,
-    "rewrite6/blocks": 1350000,
-    "rewrite6/targets": 2213124309,
-    "rewrite6/input_tokens": 2215550812
+    "climbmix/blocks": 10980000,
+    "climbmix/targets": 18354331663,
+    "climbmix/input_tokens": 18355495227,
+    "ultra_qa/blocks": 4117500,
+    "ultra_qa/targets": 6717340650,
+    "ultra_qa/input_tokens": 6717353602,
+    "cortex/blocks": 4117500,
+    "cortex/targets": 156645103,
+    "cortex/input_tokens": 6545322017,
+    "ultra_style/blocks": 4117500,
+    "ultra_style/targets": 6513296217,
+    "ultra_style/input_tokens": 6513637740,
+    "openmath/blocks": 2745000,
+    "openmath/targets": 4241678751,
+    "openmath/input_tokens": 4243092717,
+    "rewrite6/blocks": 1372500,
+    "rewrite6/targets": 2250039094,
+    "rewrite6/input_tokens": 2252505932
   },
   "sources": {
     "climbmix": {
-      "page": 1479,
-      "row": 738,
+      "page": 1504,
+      "row": 337,
       "start": {
-        "file": 71,
+        "file": 72,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 49,
+          "row_group": 66,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 6058722
+      "offset": 12288,
+      "documents": 6160721
     },
     "rewrite6": {
-      "page": 41,
-      "row": 2526,
+      "page": 5,
+      "row": 1973,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 167936
+          "row": 20480
         }
       },
-      "epoch": 16,
-      "offset": 0,
-      "documents": 3456926
+      "epoch": 17,
+      "offset": 2048,
+      "documents": 3514321
     },
     "openmath": {
-      "page": 924,
-      "row": 2465,
+      "page": 939,
+      "row": 3990,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 291,
-          "row": 504
+          "row_group": 352,
+          "row": 944
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 3787169
+      "documents": 3850134
     },
     "ultra_qa": {
-      "page": 399,
-      "row": 2747,
+      "page": 406,
+      "row": 1338,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 10712
+          "row": 39384
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1637051
+      "documents": 1664314
     },
     "ultra_style": {
-      "page": 610,
-      "row": 401,
+      "page": 620,
+      "row": 994,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 18011
+          "row": 58971
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 2498961
+      "offset": 2048,
+      "documents": 2540514
     },
     "cortex": {
-      "page": 78549,
-      "row": 103,
+      "page": 79857,
+      "row": 83,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 10054375
+      "documents": 10221779
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 300000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6492e2662f94faa6a2e8ee1d628fd698c35d02f34d6f8f370f9d4db12434c3dc", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "fbe139ff1c427ab9b5843d95ec3ecec658133f721c1f794977096107acee4acb", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "f89f82c3f4c44a52fcd532346ea71580902d757e59e1961e9b2bed0030259dba", "shard_scheduler.json": "a2c401bec97cd3feb72edade9a3ae615ffe06fc28d4766333eb54ffc7af1ddd7", "coverage.json": "251945ddb781b0d2536a5c2056932a9ec606d4b451ee97c2b33e28f6ace29956", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000300000"}
+{"step": 305000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "89b7e1e53c78bd469c045acaa45bf2613b245ad7c172ee689268c146802eaa57", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "124b35ad649617fdf90d8ce9db8b59458f1bc588be0bbe9ced8a83035f7877a9", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "f89f82c3f4c44a52fcd532346ea71580902d757e59e1961e9b2bed0030259dba", "shard_scheduler.json": "9022b90c7229c4a1b45c4f0abd2854b2ee887f5b962134d4b9e98bb6007410cc", "coverage.json": "156ed6483cb0396d79e81df2dfaa6ef016a16020c5c9fcc2a445e0bcff68144f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000305000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6492e2662f94faa6a2e8ee1d628fd698c35d02f34d6f8f370f9d4db12434c3dc
+oid sha256:89b7e1e53c78bd469c045acaa45bf2613b245ad7c172ee689268c146802eaa57
 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": 75120.45760388167,
-      "required_pages_per_second_upper_estimate": 0.006511250874045306
+      "required_token_positions_per_second": 74789.22219739005,
+      "required_pages_per_second_upper_estimate": 0.0064825402285190745
     },
     {
       "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": 9390.05720048521,
-      "required_pages_per_second_upper_estimate": 0.0037458342111397836
+      "required_token_positions_per_second": 9348.652774673756,
+      "required_pages_per_second_upper_estimate": 0.0037293173666322625
     },
     {
       "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": 18780.11440097042,
-      "required_pages_per_second_upper_estimate": 0.004344448947939665
+      "required_token_positions_per_second": 18697.30554934751,
+      "required_pages_per_second_upper_estimate": 0.004325292577502718
     },
     {
       "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": 28170.17160145563,
-      "required_pages_per_second_upper_estimate": 0.0017154156530606451
+      "required_token_positions_per_second": 28045.958324021267,
+      "required_pages_per_second_upper_estimate": 0.0017078517161615934
     },
     {
       "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": 28170.17160145563,
-      "required_pages_per_second_upper_estimate": 0.0021686177899059907
+      "required_token_positions_per_second": 28045.958324021267,
+      "required_pages_per_second_upper_estimate": 0.0021590555079647345
     },
     {
       "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": 28170.17160145563,
-      "required_pages_per_second_upper_estimate": 0.3497792517905513
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 28045.958324021267,
+      "required_pages_per_second_upper_estimate": 0.3482369385999139
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,412 +155,412 @@
   "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": 690670,
-    "producer_batch_seconds": 0.20651756988294778,
-    "producer_p95_seconds": 0.31602960099917254,
-    "gpu_update_seconds": 0.9814636698404546,
-    "producer_batches_per_second": 4.842203017238633,
-    "gpu_batches_per_second": 1.0188864149832042,
-    "producer_headroom": 4.752446343411551,
+    "producer_batch_seconds": 0.42701781460147004,
+    "producer_p95_seconds": 0.3386782650013629,
+    "gpu_update_seconds": 0.9858104929265185,
+    "producer_batches_per_second": 2.341822672979783,
+    "gpu_batches_per_second": 1.0143937472519267,
+    "producer_headroom": 2.308593363596697,
     "sustainable": true,
-    "buffer_seconds": 60.85074753010819,
+    "buffer_seconds": 60.13444006851763,
     "buffer_drain_seconds": null,
-    "shard_target_mib": 32,
-    "upload_pages_per_commit": 164,
+    "shard_target_mib": 16,
+    "upload_pages_per_commit": 76,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 628791.9512074635,
-    "upload_bytes_per_second": 2651314.957309086,
+    "generation_bytes_per_second": 297301.97635776654,
+    "upload_bytes_per_second": 5711591.078716606,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 67.4620576008221,
+    "estimated_full_shards_per_hour": 63.79408329057453,
     "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": 1.3497758060002525,
-      "bytes": 7313270,
-      "pages": 37,
+      "seconds": 1.178703661000327,
+      "bytes": 7809750,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.170668664999539,
-      "bytes": 6087923,
-      "pages": 35,
+      "seconds": 0.9307224249969295,
+      "bytes": 6774165,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.1090394350030692,
-      "bytes": 4744304,
-      "pages": 27,
+      "seconds": 0.8456139500012796,
+      "bytes": 6033435,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.304362898998079,
-      "bytes": 7978674,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.161550510001689,
+      "bytes": 11437647,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.0655079770003795,
-      "bytes": 7055873,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.8255374049986131,
+      "bytes": 1098294,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.268422623998049,
-      "bytes": 5726191,
-      "pages": 20,
+      "seconds": 0.9614813039988803,
+      "bytes": 5875069,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.1066016949989717,
-      "bytes": 7887477,
-      "pages": 44,
-      "shards": 3
+      "seconds": 1.3580094259996258,
+      "bytes": 7460093,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.019014984001842,
-      "bytes": 6005453,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.418776655002148,
+      "bytes": 10393041,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.00243121099993,
-      "bytes": 7452907,
-      "pages": 29,
+      "seconds": 1.2120838849987194,
+      "bytes": 7800195,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9669066969981941,
-      "bytes": 1217353,
-      "pages": 34,
+      "seconds": 0.7076373689997126,
+      "bytes": 1197583,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.4986801279992505,
-      "bytes": 15911648,
-      "pages": 32,
-      "shards": 4
+      "seconds": 1.2053027080000902,
+      "bytes": 10494285,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.1255449119998957,
-      "bytes": 2547314,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.2219008450010733,
+      "bytes": 7585902,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2029692989999603,
+      "bytes": 1075789,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.1576964999985648,
-      "bytes": 6153703,
+      "seconds": 1.6663710430002538,
+      "bytes": 11892665,
       "pages": 36,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.0455056530008733,
-      "bytes": 6073870,
-      "pages": 35,
+      "seconds": 1.4550072420024662,
+      "bytes": 2481838,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.0066101270022045,
-      "bytes": 7063346,
-      "pages": 30,
+      "seconds": 1.1862465560006967,
+      "bytes": 10217288,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.161024010998517,
-      "bytes": 6075989,
-      "pages": 36,
+      "seconds": 1.3918344490011805,
+      "bytes": 7389964,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.8424355740025931,
-      "bytes": 669079,
-      "pages": 20,
+      "seconds": 0.8301882180021494,
+      "bytes": 1174116,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.4896883629990043,
-      "bytes": 13505965,
-      "pages": 48,
-      "shards": 4
+      "seconds": 1.4423049569995783,
+      "bytes": 10949179,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.098302499998681,
-      "bytes": 6017424,
-      "pages": 23,
-      "shards": 2
+      "seconds": 1.3191998570000578,
+      "bytes": 13241805,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.0887724580024951,
-      "bytes": 6603634,
-      "pages": 44,
-      "shards": 3
+      "seconds": 1.0903980459988816,
+      "bytes": 954449,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.2179049539990956,
-      "bytes": 6068321,
+      "seconds": 1.1526671549981984,
+      "bytes": 6176156,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 2.3319496820004133,
-      "bytes": 5855101,
-      "pages": 27,
+      "seconds": 1.0391054690007877,
+      "bytes": 2626145,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5211132250005903,
-      "bytes": 6886706,
-      "pages": 29,
+      "seconds": 1.0629035870006192,
+      "bytes": 10213495,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.7056977829997777,
-      "bytes": 6292570,
-      "pages": 16,
-      "shards": 2
-    },
-    {
-      "seconds": 10.767963721002161,
-      "bytes": 5800923,
+      "seconds": 0.9228753829993366,
+      "bytes": 2415467,
       "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 6.862793137002882,
-      "bytes": 7606529,
-      "pages": 38,
+      "seconds": 1.343091220001952,
+      "bytes": 11741900,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 3.3253811320028035,
-      "bytes": 1169016,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2118571440005326,
+      "bytes": 6204328,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 12.285631268001453,
-      "bytes": 12643645,
-      "pages": 42,
-      "shards": 4
+      "seconds": 1.1613783780012454,
+      "bytes": 6337641,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 2.9626744119996147,
-      "bytes": 6098299,
-      "pages": 35,
+      "seconds": 1.4139873040003295,
+      "bytes": 5988516,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1153004340012558,
-      "bytes": 6713918,
-      "pages": 32,
-      "shards": 2
+      "seconds": 1.5244087760002003,
+      "bytes": 914665,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 5.8015411799970025,
-      "bytes": 7310989,
-      "pages": 32,
+      "seconds": 1.1577779409999494,
+      "bytes": 7413917,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 3.285385278999456,
-      "bytes": 842975,
-      "pages": 25,
-      "shards": 1
-    },
-    {
-      "seconds": 2.201649025999359,
-      "bytes": 9943938,
+      "seconds": 1.1579117699984636,
+      "bytes": 10571255,
       "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.3160915080006816,
-      "bytes": 2528581,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.0644437480004854,
+      "bytes": 8538245,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 8.146259255998302,
-      "bytes": 6143251,
+      "seconds": 1.3977851220006414,
+      "bytes": 6073631,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 6.752744052999333,
-      "bytes": 5959933,
+      "seconds": 1.1060321420009132,
+      "bytes": 6015087,
       "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 8.626221913000336,
-      "bytes": 11949436,
-      "pages": 25,
-      "shards": 4
+      "seconds": 1.2197453389999282,
+      "bytes": 6685693,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 13.160608485999546,
-      "bytes": 6360120,
-      "pages": 35,
+      "seconds": 1.2612095350013988,
+      "bytes": 6299046,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 7.225236703998235,
-      "bytes": 1179968,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.8313243659977161,
+      "bytes": 944700,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 4.945576189998974,
-      "bytes": 7777829,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.2853576430024987,
+      "bytes": 13551063,
+      "pages": 43,
+      "shards": 4
     },
     {
-      "seconds": 1.2244086760001665,
-      "bytes": 5469696,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.137746813998092,
+      "bytes": 10438351,
+      "pages": 27,
+      "shards": 3
     },
     {
-      "seconds": 1.178703661000327,
-      "bytes": 7809750,
-      "pages": 38,
-      "shards": 3
+      "seconds": 0.9337895320022653,
+      "bytes": 2628632,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 0.9307224249969295,
-      "bytes": 6774165,
-      "pages": 31,
+      "seconds": 0.9628497100020468,
+      "bytes": 5747199,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.8456139500012796,
-      "bytes": 6033435,
+      "seconds": 0.9260915579980065,
+      "bytes": 968750,
       "pages": 30,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.161550510001689,
-      "bytes": 11437647,
-      "pages": 35,
+      "seconds": 1.5047391669977515,
+      "bytes": 11683531,
+      "pages": 33,
       "shards": 4
     },
     {
-      "seconds": 0.8255374049986131,
-      "bytes": 1098294,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2526458680003998,
+      "bytes": 12133817,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 0.9614813039988803,
-      "bytes": 5875069,
-      "pages": 29,
+      "seconds": 0.8809280429995852,
+      "bytes": 2767942,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3580094259996258,
-      "bytes": 7460093,
-      "pages": 38,
-      "shards": 4
+      "seconds": 1.0177543880017765,
+      "bytes": 5824791,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.418776655002148,
-      "bytes": 10393041,
-      "pages": 31,
+      "seconds": 1.107241077999788,
+      "bytes": 9747301,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.2120838849987194,
-      "bytes": 7800195,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.9869726110009651,
+      "bytes": 2644359,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.7076373689997126,
-      "bytes": 1197583,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3357893329994113,
+      "bytes": 5825933,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.2053027080000902,
-      "bytes": 10494285,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.0486931550003646,
+      "bytes": 6833428,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.2219008450010733,
-      "bytes": 7585902,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.242159831999743,
+      "bytes": 12337974,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.2029692989999603,
-      "bytes": 1075789,
-      "pages": 31,
-      "shards": 1
+      "seconds": 1.3698127219977323,
+      "bytes": 6046909,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.6663710430002538,
-      "bytes": 11892665,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.4521962930011796,
+      "bytes": 2626843,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.4550072420024662,
-      "bytes": 2481838,
-      "pages": 31,
+      "seconds": 1.2158071900012146,
+      "bytes": 5739703,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.1862465560006967,
-      "bytes": 10217288,
+      "seconds": 1.3039640329989197,
+      "bytes": 9738932,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.3918344490011805,
-      "bytes": 7389964,
-      "pages": 31,
+      "seconds": 1.1697420529999363,
+      "bytes": 8456264,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 0.8301882180021494,
-      "bytes": 1174116,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.2002514119994885,
+      "bytes": 5878811,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.4423049569995783,
-      "bytes": 10949179,
+      "seconds": 1.0716100809986528,
+      "bytes": 2567688,
       "pages": 36,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.3191998570000578,
-      "bytes": 13241805,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.52958795199811,
+      "bytes": 10493875,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.0903980459988816,
-      "bytes": 954449,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.2153502640030638,
+      "bytes": 6178782,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.1526671549981984,
-      "bytes": 6176156,
-      "pages": 36,
+      "seconds": 1.1377685129991733,
+      "bytes": 2613014,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0391054690007877,
-      "bytes": 2626145,
-      "pages": 36,
+      "seconds": 1.2625757009991503,
+      "bytes": 11728601,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9978417950005678,
+      "bytes": 5753379,
+      "pages": 30,
       "shards": 2
     }
   ],

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fbe139ff1c427ab9b5843d95ec3ecec658133f721c1f794977096107acee4acb
-size 721764543
+oid sha256:124b35ad649617fdf90d8ce9db8b59458f1bc588be0bbe9ced8a83035f7877a9
+size 721760127

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,