Qwen90M step 875000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 870000,
+  "step": 875000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b",
+    "model.safetensors": "9f61ed48f2cb4c3f7a06a34a94d9060f253db6eb2c73e5fe5e4c3640c5f78d50",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed",
-    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
+    "trainer.pt": "a291745e8cb415b9662fa7db6054a73b2972d734926e2a0eeb0e4f105d90a9b2",
+    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
     "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367",
-    "shard_scheduler.json": "328e638e7ad91d39ab501d496c26d65bbb1a3067cdf2542ba84f30916f4660be",
-    "coverage.json": "7d047105e0eebd85909b8ec68f55ac31224c6dec9049175243e510c634cd4bd3",
+    "shard_scheduler.json": "4d5b87679412b97e1700b3b9d9840dca98fd012ebe6d7c8389acbcbf734e9db5",
+    "coverage.json": "54ea23c51a1da54ebc5dce287846cffb5ef7a00a7d2bf8f7eed2277d17896c85",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 870000,
+  "step": 875000,
   "stats": {
-    "climbmix/blocks": 31320000,
-    "climbmix/targets": 52356514020,
-    "climbmix/input_tokens": 52359831944,
-    "ultra_qa/blocks": 11745000,
-    "ultra_qa/targets": 19161264699,
-    "ultra_qa/input_tokens": 19161301550,
-    "cortex/blocks": 11745000,
-    "cortex/targets": 446591895,
-    "cortex/input_tokens": 18669449496,
-    "ultra_style/blocks": 11745000,
-    "ultra_style/targets": 18577453797,
-    "ultra_style/input_tokens": 18578427724,
-    "openmath/blocks": 7830000,
-    "openmath/targets": 12098676149,
-    "openmath/input_tokens": 12102707529,
-    "rewrite6/blocks": 3915000,
-    "rewrite6/targets": 6418117328,
-    "rewrite6/input_tokens": 6425153910
+    "climbmix/blocks": 31500000,
+    "climbmix/targets": 52657150038,
+    "climbmix/input_tokens": 52660486950,
+    "ultra_qa/blocks": 11812500,
+    "ultra_qa/targets": 19271353866,
+    "ultra_qa/input_tokens": 19271390893,
+    "cortex/blocks": 11812500,
+    "cortex/targets": 449143762,
+    "cortex/input_tokens": 18776747701,
+    "ultra_style/blocks": 11812500,
+    "ultra_style/targets": 18684178798,
+    "ultra_style/input_tokens": 18685158198,
+    "openmath/blocks": 7875000,
+    "openmath/targets": 12168335207,
+    "openmath/input_tokens": 12172389842,
+    "rewrite6/blocks": 3937500,
+    "rewrite6/targets": 6455009145,
+    "rewrite6/input_tokens": 6462085998
   },
   "sources": {
     "climbmix": {
-      "page": 4289,
-      "row": 1766,
+      "page": 4314,
+      "row": 579,
       "start": {
-        "file": 207,
+        "file": 208,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 51,
+          "row_group": 68,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 17569510
+      "offset": 2048,
+      "documents": 17670723
     },
     "rewrite6": {
-      "page": 40,
-      "row": 1476,
+      "page": 4,
+      "row": 876,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 163840
+          "row": 16384
         }
       },
-      "epoch": 48,
-      "offset": 0,
-      "documents": 10024708
+      "epoch": 49,
+      "offset": 2048,
+      "documents": 10082056
     },
     "openmath": {
-      "page": 2680,
-      "row": 3506,
+      "page": 2696,
+      "row": 1147,
       "start": {
         "file": 25,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 61,
-          "row": 32
+          "row_group": 126,
+          "row": 568
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 10980786
+      "documents": 11043963
     },
     "ultra_qa": {
-      "page": 1159,
-      "row": 365,
+      "page": 1165,
+      "row": 3053,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 6079
+          "row": 30655
         }
       },
       "epoch": 0,
       "offset": 4096,
-      "documents": 4747629
+      "documents": 4774893
     },
     "ultra_style": {
-      "page": 1768,
-      "row": 3754,
+      "page": 1779,
+      "row": 128,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
-          "row": 53002
+          "row_group": 5,
+          "row": 12636
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 7245482
+      "documents": 7286912
     },
     "cortex": {
-      "page": 227777,
-      "row": 116,
+      "page": 229085,
+      "row": 121,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 29155572
+      "documents": 29323001
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 870000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "328e638e7ad91d39ab501d496c26d65bbb1a3067cdf2542ba84f30916f4660be", "coverage.json": "7d047105e0eebd85909b8ec68f55ac31224c6dec9049175243e510c634cd4bd3", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000870000"}
+{"step": 875000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "9f61ed48f2cb4c3f7a06a34a94d9060f253db6eb2c73e5fe5e4c3640c5f78d50", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a291745e8cb415b9662fa7db6054a73b2972d734926e2a0eeb0e4f105d90a9b2", "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "4d5b87679412b97e1700b3b9d9840dca98fd012ebe6d7c8389acbcbf734e9db5", "coverage.json": "54ea23c51a1da54ebc5dce287846cffb5ef7a00a7d2bf8f7eed2277d17896c85", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000875000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b
+oid sha256:9f61ed48f2cb4c3f7a06a34a94d9060f253db6eb2c73e5fe5e4c3640c5f78d50
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 77384.34908444676,
-      "required_pages_per_second_upper_estimate": 0.005353628826089951
+      "required_token_positions_per_second": 75597.86980031131,
+      "required_pages_per_second_upper_estimate": 0.005230036044010423
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9673.043635555845,
-      "required_pages_per_second_upper_estimate": 0.004049901374086377
+      "required_token_positions_per_second": 9449.733725038914,
+      "required_pages_per_second_upper_estimate": 0.003956406177794145
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19346.08727111169,
-      "required_pages_per_second_upper_estimate": 0.004029279787942392
+      "required_token_positions_per_second": 18899.467450077827,
+      "required_pages_per_second_upper_estimate": 0.0039362606573777655
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29019.130906667535,
-      "required_pages_per_second_upper_estimate": 0.001739652588207436
+      "required_token_positions_per_second": 28349.201175116745,
+      "required_pages_per_second_upper_estimate": 0.001699491323724437
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29019.130906667535,
-      "required_pages_per_second_upper_estimate": 0.0022561097863261297
+      "required_token_positions_per_second": 28349.201175116745,
+      "required_pages_per_second_upper_estimate": 0.0022040256964075268
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.04061180899998362,
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
-      "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 29019.130906667535,
-      "required_pages_per_second_upper_estimate": 0.37222625295554873
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28349.201175116745,
+      "required_pages_per_second_upper_estimate": 0.3636331136737182
     }
   ],
   "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": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.10117103117187298,
-    "producer_p95_seconds": 0.2711704980029026,
-    "gpu_update_seconds": 0.9527507935686489,
-    "producer_batches_per_second": 9.884252324177304,
-    "gpu_batches_per_second": 1.049592408371945,
-    "producer_headroom": 9.417229245692688,
+    "producer_batch_seconds": 0.12715467205484288,
+    "producer_p95_seconds": 0.31598223599939956,
+    "gpu_update_seconds": 0.9752655755347274,
+    "producer_batches_per_second": 7.864437726430466,
+    "gpu_batches_per_second": 1.0253617323175905,
+    "producer_headroom": 7.669915385524231,
     "sustainable": true,
-    "buffer_seconds": 60.02329999482488,
+    "buffer_seconds": 60.466465683153096,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 86,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 322491.0104688514,
-    "upload_bytes_per_second": 4817943.822503319,
+    "generation_bytes_per_second": 484630.29120586516,
+    "upload_bytes_per_second": 4285666.65230903,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 69.1990636401096,
+    "estimated_full_shards_per_hour": 103.99037887699095,
     "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": 0.990031227000145,
-      "bytes": 1215408,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.6203178860014305,
+      "bytes": 6721657,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 1.4002689310000278,
-      "bytes": 11880783,
+      "seconds": 1.8514023890020326,
+      "bytes": 12113021,
       "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 1.1156020849994093,
-      "bytes": 6256874,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.0979482909970102,
+      "bytes": 1151793,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.0478186670006835,
-      "bytes": 6856744,
-      "pages": 36,
+      "seconds": 1.1090724049972778,
+      "bytes": 6289086,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.187728501001402,
-      "bytes": 7914573,
-      "pages": 30,
+      "seconds": 1.7820321440012776,
+      "bytes": 7365760,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.098511371998029,
-      "bytes": 5354193,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.5119859270016605,
+      "bytes": 1025135,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 2.7016539569995075,
-      "bytes": 7555720,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.464499204997992,
+      "bytes": 17226254,
+      "pages": 41,
+      "shards": 5
     },
     {
-      "seconds": 1.483174205000978,
-      "bytes": 5900678,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.9209507179984939,
+      "bytes": 949705,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 0.8595360259969311,
-      "bytes": 1182793,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.9317821920012648,
+      "bytes": 6082541,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.7856543759990018,
-      "bytes": 17299146,
-      "pages": 27,
-      "shards": 6
+      "seconds": 1.9298742110004241,
+      "bytes": 11871053,
+      "pages": 33,
+      "shards": 4
     },
     {
-      "seconds": 0.9336543670033279,
-      "bytes": 1145946,
+      "seconds": 1.752713240999583,
+      "bytes": 1247797,
       "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.1468644119995588,
-      "bytes": 6284755,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.4635059559986985,
+      "bytes": 7719999,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.0925358679996862,
-      "bytes": 2671177,
-      "pages": 37,
+      "seconds": 1.2099752760004776,
+      "bytes": 6423731,
+      "pages": 22,
       "shards": 2
     },
     {
-      "seconds": 1.42488349499763,
-      "bytes": 9837963,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2818151600004057,
-      "bytes": 7876217,
+      "seconds": 1.4774333919995115,
+      "bytes": 10519718,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1551925370004028,
-      "bytes": 6771836,
+      "seconds": 1.5275958259990148,
+      "bytes": 7252352,
       "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.460754736999661,
-      "bytes": 10771768,
-      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.9620606520002184,
-      "bytes": 2617341,
+      "seconds": 1.0232120440014114,
+      "bytes": 1138613,
       "pages": 35,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.1643879789990024,
-      "bytes": 6176643,
-      "pages": 37,
+      "seconds": 1.0960563930020726,
+      "bytes": 6111629,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.7241964509994432,
-      "bytes": 5999571,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.3281459399986488,
+      "bytes": 11209095,
+      "pages": 33,
+      "shards": 4
     },
     {
-      "seconds": 1.129534389001492,
-      "bytes": 2663790,
-      "pages": 37,
+      "seconds": 1.3028533390024677,
+      "bytes": 6867913,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.8240933249981026,
-      "bytes": 16162613,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 1.3452121949994762,
-      "bytes": 2460437,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.275760786000319,
+      "bytes": 7467840,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.3531316250009695,
-      "bytes": 6148520,
+      "seconds": 1.235418092001055,
+      "bytes": 6152144,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9429252529989753,
-      "bytes": 341818,
-      "pages": 9,
-      "shards": 1
-    },
-    {
-      "seconds": 1.6665152259993192,
-      "bytes": 10630836,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.3146963539984426,
+      "bytes": 5075870,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.7391449930000817,
-      "bytes": 7296963,
-      "pages": 39,
+      "seconds": 1.278188054999191,
+      "bytes": 7607915,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.3247330819976924,
-      "bytes": 862254,
+      "seconds": 1.058052360000147,
+      "bytes": 885421,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.7279805650032358,
-      "bytes": 13248087,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.6197210989994346,
+      "bytes": 11971576,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.2414953469997272,
-      "bytes": 4717232,
-      "pages": 37,
-      "shards": 2
+      "seconds": 3.1373434490014915,
+      "bytes": 11875917,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.1668012100017222,
-      "bytes": 6253390,
-      "pages": 32,
-      "shards": 2
+      "seconds": 1.013686755002709,
+      "bytes": 986186,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.4145583839999745,
-      "bytes": 7569385,
+      "seconds": 1.333218751999084,
+      "bytes": 7645187,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1982823690013902,
-      "bytes": 1172595,
+      "seconds": 1.3886132959996758,
+      "bytes": 6063422,
       "pages": 34,
-      "shards": 1
+      "shards": 2
     },
     {
-      "seconds": 1.8298632039986842,
-      "bytes": 11926092,
-      "pages": 29,
+      "seconds": 1.2067982320004376,
+      "bytes": 5015589,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4427611729988712,
+      "bytes": 13297372,
+      "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 1.4296780900003796,
-      "bytes": 6878310,
-      "pages": 32,
+      "seconds": 1.798427575002279,
+      "bytes": 1107567,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.173846097997739,
-      "bytes": 6578162,
-      "pages": 36,
+      "seconds": 1.1358039750011812,
+      "bytes": 5944518,
+      "pages": 25,
       "shards": 2
     },
     {
-      "seconds": 1.273192676002509,
-      "bytes": 7543138,
-      "pages": 40,
-      "shards": 3
+      "seconds": 1.3701994980001473,
+      "bytes": 11970289,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.0804058459980297,
-      "bytes": 5227494,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.0856028650014196,
+      "bytes": 1171092,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.174932928999624,
-      "bytes": 7282353,
+      "seconds": 3.5694688390030933,
+      "bytes": 7359997,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.3665898539984482,
-      "bytes": 6068190,
-      "pages": 36,
+      "seconds": 1.4348496300008264,
+      "bytes": 6710244,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.6203178860014305,
-      "bytes": 6721657,
-      "pages": 29,
+      "seconds": 1.629695798998,
+      "bytes": 10655327,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.8514023890020326,
-      "bytes": 12113021,
+      "seconds": 1.3922439659982047,
+      "bytes": 7460601,
       "pages": 32,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 1.0979482909970102,
-      "bytes": 1151793,
+      "seconds": 1.1743250350009475,
+      "bytes": 1256562,
       "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.1090724049972778,
-      "bytes": 6289086,
-      "pages": 37,
+      "seconds": 1.874451825999131,
+      "bytes": 7588108,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "seconds": 1.605746043002,
+      "bytes": 4717702,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.7820321440012776,
-      "bytes": 7365760,
-      "pages": 31,
+      "seconds": 1.3987950150003599,
+      "bytes": 11634070,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.5119859270016605,
-      "bytes": 1025135,
-      "pages": 32,
+      "seconds": 1.6460893040020892,
+      "bytes": 7447219,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9642833869984315,
+      "bytes": 493177,
+      "pages": 16,
       "shards": 1
     },
     {
-      "seconds": 1.464499204997992,
-      "bytes": 17226254,
-      "pages": 41,
-      "shards": 5
+      "seconds": 1.69654296200315,
+      "bytes": 11528790,
+      "pages": 47,
+      "shards": 3
     },
     {
-      "seconds": 0.9209507179984939,
-      "bytes": 949705,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.0500477669993415,
+      "bytes": 2444431,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.9317821920012648,
-      "bytes": 6082541,
-      "pages": 37,
+      "seconds": 1.1570196089996898,
+      "bytes": 6331951,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.9298742110004241,
-      "bytes": 11871053,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.2861647640020237,
+      "bytes": 6463355,
+      "pages": 21,
+      "shards": 2
     },
     {
-      "seconds": 1.752713240999583,
-      "bytes": 1247797,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.3757250710004882,
+      "bytes": 11835703,
+      "pages": 44,
+      "shards": 4
     },
     {
-      "seconds": 1.4635059559986985,
-      "bytes": 7719999,
-      "pages": 40,
-      "shards": 3
+      "seconds": 1.5062627519982925,
+      "bytes": 6228251,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.2099752760004776,
-      "bytes": 6423731,
-      "pages": 22,
+      "seconds": 1.0129257990010956,
+      "bytes": 2288801,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.4774333919995115,
-      "bytes": 10519718,
+      "seconds": 1.3661311769974418,
+      "bytes": 11305021,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.5275958259990148,
-      "bytes": 7252352,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.2991957800004457,
+      "bytes": 5765469,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.0232120440014114,
-      "bytes": 1138613,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.2353659159998642,
+      "bytes": 8486468,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.0960563930020726,
-      "bytes": 6111629,
-      "pages": 38,
+      "seconds": 1.681274881997524,
+      "bytes": 6344982,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.3281459399986488,
-      "bytes": 11209095,
-      "pages": 33,
-      "shards": 4
-    },
-    {
-      "seconds": 1.3028533390024677,
-      "bytes": 6867913,
+      "seconds": 1.540335444999073,
+      "bytes": 2545551,
       "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.275760786000319,
-      "bytes": 7467840,
-      "pages": 35,
+      "seconds": 1.1612642559994129,
+      "bytes": 9600274,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.235418092001055,
-      "bytes": 6152144,
-      "pages": 36,
+      "seconds": 1.406910452999,
+      "bytes": 5890041,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.3146963539984426,
-      "bytes": 5075870,
-      "pages": 30,
+      "seconds": 1.0961665050017473,
+      "bytes": 2428850,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.278188054999191,
-      "bytes": 7607915,
-      "pages": 40,
+      "seconds": 6.228057981999882,
+      "bytes": 11900976,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.058052360000147,
-      "bytes": 885421,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.5676271939992148,
+      "bytes": 12360812,
+      "pages": 44,
+      "shards": 4
+    },
+    {
+      "seconds": 1.7888847590002115,
+      "bytes": 1223080,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 4.01536119299999,
+      "bytes": 5846921,
+      "pages": 29,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed
-size 721769279
+oid sha256:a291745e8cb415b9662fa7db6054a73b2972d734926e2a0eeb0e4f105d90a9b2
+size 721771711

training_config.json

@@ -53,7 +53,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
+  "prefetch_batches": 62,
   "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,