Qwen90M step 170000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 165000,
+  "step": 170000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "2055e13f7b5b304d0c2206d7fb9fd41c1dbf127325878ef730f5b741d6555462",
+    "model.safetensors": "eae93b5a987696907c76cdf5c39f1ece9ef2dc6c0dacd54567e9b61a70909330",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "4aadd2411e27d45ea0852b13570a9cc811de9a400c6b05705bce64c357639250",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "68b2bd34b599c12b63f5eb6b5156f7905dc9454684b6f321753aaa307c958f77",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
-    "shard_scheduler.json": "37c2d3b415e95706fcd09442dbfed95fbe0847366375512b565a88834bde0eeb",
-    "coverage.json": "6b16c159bcce254a2303d601ede22d8e32b19a35679351555cb18cba36d7f12d",
+    "shard_scheduler.json": "6b80239d2fad6471bd75c8996f07573047714dceff819ccf239934bc4693936c",
+    "coverage.json": "33a1ecc073a8506314c0c7e8b1455eaacfff9bce622ac5436c9d5ec2fcbfbf5d",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 165000,
+  "step": 170000,
   "stats": {
-    "climbmix/blocks": 5940000,
-    "climbmix/targets": 9927628187,
-    "climbmix/input_tokens": 9928257678,
-    "ultra_qa/blocks": 2227500,
-    "ultra_qa/targets": 3633811973,
-    "ultra_qa/input_tokens": 3633818963,
-    "cortex/blocks": 2227500,
-    "cortex/targets": 84725586,
-    "cortex/input_tokens": 3540651120,
-    "ultra_style/blocks": 2227500,
-    "ultra_style/targets": 3523134670,
-    "ultra_style/input_tokens": 3523319504,
-    "openmath/blocks": 1485000,
-    "openmath/targets": 2294270266,
-    "openmath/input_tokens": 2295034673,
-    "rewrite6/blocks": 742500,
-    "rewrite6/targets": 1217215870,
-    "rewrite6/input_tokens": 1218550219
+    "climbmix/blocks": 6120000,
+    "climbmix/targets": 10228213717,
+    "climbmix/input_tokens": 10228862440,
+    "ultra_qa/blocks": 2295000,
+    "ultra_qa/targets": 3743992865,
+    "ultra_qa/input_tokens": 3744000053,
+    "cortex/blocks": 2295000,
+    "cortex/targets": 87303716,
+    "cortex/input_tokens": 3647892000,
+    "ultra_style/blocks": 2295000,
+    "ultra_style/targets": 3630136904,
+    "ultra_style/input_tokens": 3630327119,
+    "openmath/blocks": 1530000,
+    "openmath/targets": 2363841526,
+    "openmath/input_tokens": 2364629111,
+    "rewrite6/blocks": 765000,
+    "rewrite6/targets": 1254067034,
+    "rewrite6/input_tokens": 1255442274
   },
   "sources": {
     "climbmix": {
-      "page": 814,
-      "row": 242,
+      "page": 838,
+      "row": 3535,
       "start": {
-        "file": 39,
+        "file": 40,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 27,
+          "row_group": 40,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 3334386
+      "documents": 3435983
     },
     "rewrite6": {
-      "page": 12,
-      "row": 3197,
+      "page": 26,
+      "row": 4050,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 49152
+          "row": 106496
         }
       },
       "epoch": 9,
-      "offset": 0,
-      "documents": 1900985
+      "offset": 2048,
+      "documents": 1959182
     },
     "openmath": {
-      "page": 508,
-      "row": 1788,
+      "page": 523,
+      "row": 3399,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 334,
-          "row": 168
+          "row_group": 395,
+          "row": 608
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 2082556
+      "offset": 2048,
+      "documents": 2145607
     },
     "ultra_qa": {
-      "page": 219,
-      "row": 3510,
+      "page": 226,
+      "row": 2070,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 54490
+          "row_group": 6,
+          "row": 18609
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 900534
+      "offset": 4096,
+      "documents": 927766
     },
     "ultra_style": {
-      "page": 335,
-      "row": 2292,
+      "page": 345,
+      "row": 2329,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 2278
+          "row": 43238
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1374452
+      "documents": 1415449
     },
     "cortex": {
-      "page": 43197,
-      "row": 83,
+      "page": 44505,
+      "row": 35,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 5529299
+      "documents": 5696675
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 165000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2055e13f7b5b304d0c2206d7fb9fd41c1dbf127325878ef730f5b741d6555462", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "4aadd2411e27d45ea0852b13570a9cc811de9a400c6b05705bce64c357639250", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "37c2d3b415e95706fcd09442dbfed95fbe0847366375512b565a88834bde0eeb", "coverage.json": "6b16c159bcce254a2303d601ede22d8e32b19a35679351555cb18cba36d7f12d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000165000"}
+{"step": 170000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "eae93b5a987696907c76cdf5c39f1ece9ef2dc6c0dacd54567e9b61a70909330", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "68b2bd34b599c12b63f5eb6b5156f7905dc9454684b6f321753aaa307c958f77", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "6b80239d2fad6471bd75c8996f07573047714dceff819ccf239934bc4693936c", "coverage.json": "33a1ecc073a8506314c0c7e8b1455eaacfff9bce622ac5436c9d5ec2fcbfbf5d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000170000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2055e13f7b5b304d0c2206d7fb9fd41c1dbf127325878ef730f5b741d6555462
+oid sha256:eae93b5a987696907c76cdf5c39f1ece9ef2dc6c0dacd54567e9b61a70909330
 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": 76024.17600051314,
-      "required_pages_per_second_upper_estimate": 0.006589582894211986
+      "required_token_positions_per_second": 74872.19943677516,
+      "required_pages_per_second_upper_estimate": 0.0064897324853250855
     },
     {
       "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": 9503.022000064142,
-      "required_pages_per_second_upper_estimate": 0.0037908975586660853
+      "required_token_positions_per_second": 9359.024929596895,
+      "required_pages_per_second_upper_estimate": 0.0037334549743086387
     },
     {
       "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": 19006.044000128284,
-      "required_pages_per_second_upper_estimate": 0.004396713784479697
+      "required_token_positions_per_second": 18718.04985919379,
+      "required_pages_per_second_upper_estimate": 0.004330091408498271
     },
     {
       "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": 28509.06600019243,
-      "required_pages_per_second_upper_estimate": 0.0017360525439022204
+      "required_token_positions_per_second": 28077.074788790687,
+      "required_pages_per_second_upper_estimate": 0.0017097465456105776
     },
     {
       "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": 28509.06600019243,
-      "required_pages_per_second_upper_estimate": 0.002194706818840488
+      "required_token_positions_per_second": 28077.074788790687,
+      "required_pages_per_second_upper_estimate": 0.002161450939558572
     },
     {
       "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": 28509.06600019243,
-      "required_pages_per_second_upper_estimate": 0.35398718601627116
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 28077.074788790687,
+      "required_pages_per_second_upper_estimate": 0.34862330095224164
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,370 +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": 688946,
-    "producer_batch_seconds": 0.09048123760162241,
-    "producer_p95_seconds": 0.1997182420000172,
-    "gpu_update_seconds": 0.9697967656959854,
-    "producer_batches_per_second": 11.05201505314146,
-    "gpu_batches_per_second": 1.0311438802152932,
-    "producer_headroom": 10.718208452959932,
+    "producer_batch_seconds": 0.06762013767189501,
+    "producer_p95_seconds": 0.20516962799956673,
+    "gpu_update_seconds": 0.9847179668103464,
+    "producer_batches_per_second": 14.78849399645086,
+    "gpu_batches_per_second": 1.015519198089941,
+    "producer_headroom": 14.562495740372105,
     "sustainable": true,
-    "buffer_seconds": 60.127399473151094,
+    "buffer_seconds": 60.06779597543113,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 83,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 314248.84728296526,
-    "upload_bytes_per_second": 5212065.682698314,
+    "generation_bytes_per_second": 428682.5490002683,
+    "upload_bytes_per_second": 5225926.143782697,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 67.43048728815764,
+    "estimated_full_shards_per_hour": 91.98529579645191,
     "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": 1.0642098359999181,
-      "bytes": 45649,
-      "pages": 1,
+      "seconds": 1.0256425439993109,
+      "bytes": 848800,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.2323853809998582,
-      "bytes": 7278265,
+      "seconds": 1.479707024000163,
+      "bytes": 13318632,
+      "pages": 43,
+      "shards": 4
+    },
+    {
+      "seconds": 1.157918292999966,
+      "bytes": 10292312,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 2.16527700000006,
+      "bytes": 1181305,
       "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.226145381000606,
+      "bytes": 7407097,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9598245239999414,
-      "bytes": 1062466,
+      "seconds": 0.7928119780008274,
+      "bytes": 1141631,
       "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.1868087180000657,
-      "bytes": 11845509,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.3008312719994137,
+      "bytes": 10073702,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 0.9347497089997887,
-      "bytes": 6852819,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.3391879739992874,
+      "bytes": 13717263,
+      "pages": 44,
+      "shards": 4
     },
     {
-      "seconds": 1.0938459449998845,
-      "bytes": 6142145,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.882563871999082,
+      "bytes": 1071171,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.2391148739998243,
-      "bytes": 7141661,
-      "pages": 29,
+      "seconds": 1.5854626939999434,
+      "bytes": 7158059,
+      "pages": 27,
       "shards": 3
     },
     {
-      "seconds": 3.5350073540002995,
-      "bytes": 5274454,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.1073119590000715,
+      "bytes": 10577372,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.0228857370002515,
-      "bytes": 6079721,
+      "seconds": 0.7750392130001273,
+      "bytes": 1199165,
       "pages": 34,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.587895904999641,
-      "bytes": 7946846,
-      "pages": 38,
+      "seconds": 1.0709359819993551,
+      "bytes": 7553738,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.0177612440002122,
-      "bytes": 6784132,
-      "pages": 31,
+      "seconds": 0.9643511739996029,
+      "bytes": 6918272,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.3712029609996534,
-      "bytes": 12398330,
-      "pages": 38,
-      "shards": 4
+      "seconds": 1.0286668139997346,
+      "bytes": 9530498,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 0.7755766290001702,
-      "bytes": 1144894,
-      "pages": 33,
-      "shards": 1
+      "seconds": 0.9772154700003739,
+      "bytes": 2553165,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.4222684020005545,
-      "bytes": 6265294,
-      "pages": 27,
+      "seconds": 1.3266675979994034,
+      "bytes": 5992849,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.9120372190000126,
-      "bytes": 2656908,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.2236139670003467,
+      "bytes": 7766768,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.6420339440001044,
-      "bytes": 5755818,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.4731977419996838,
+      "bytes": 6241189,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.4588776000000507,
-      "bytes": 16176314,
-      "pages": 44,
-      "shards": 4
+      "seconds": 1.5922990050003136,
+      "bytes": 11693360,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 0.8323371429996769,
-      "bytes": 2088001,
-      "pages": 18,
+      "seconds": 0.8287107050000486,
+      "bytes": 2337614,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2336509919996388,
-      "bytes": 6127983,
-      "pages": 36,
+      "seconds": 0.9556833940005163,
+      "bytes": 6215473,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0918112409999594,
-      "bytes": 968274,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.3753932430008717,
+      "bytes": 9934254,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.2902597590000369,
-      "bytes": 12650540,
-      "pages": 36,
-      "shards": 4
+      "seconds": 0.9314350770000601,
+      "bytes": 2320145,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.0202606470002138,
-      "bytes": 1124000,
-      "pages": 34,
+      "seconds": 1.1146210560000327,
+      "bytes": 5994268,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.3055697850004435,
-      "bytes": 13351938,
-      "pages": 38,
-      "shards": 4
+      "seconds": 0.7972013050002715,
+      "bytes": 688709,
+      "pages": 19,
+      "shards": 1
     },
     {
-      "seconds": 1.0015427250000357,
-      "bytes": 6023702,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.845211140000174,
+      "bytes": 18359200,
+      "pages": 44,
+      "shards": 5
     },
     {
-      "seconds": 1.052357557000505,
-      "bytes": 4323709,
-      "pages": 27,
-      "shards": 2
+      "seconds": 0.8465310680003313,
+      "bytes": 1116330,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.0870202730002347,
-      "bytes": 7523830,
-      "pages": 37,
+      "seconds": 1.1592603440003586,
+      "bytes": 7563694,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8949287559998993,
-      "bytes": 6264749,
+      "seconds": 1.1315413200009061,
+      "bytes": 5996697,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0034355950001554,
-      "bytes": 860910,
-      "pages": 25,
+      "seconds": 2.1342484679989866,
+      "bytes": 1010482,
+      "pages": 32,
       "shards": 1
     },
     {
-      "seconds": 1.6520775649996722,
-      "bytes": 17392082,
+      "seconds": 1.0687636729999213,
+      "bytes": 11353142,
+      "pages": 26,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1774445949995425,
+      "bytes": 12065836,
       "pages": 43,
-      "shards": 5
+      "shards": 3
     },
     {
-      "seconds": 1.2575243370001772,
-      "bytes": 1055853,
-      "pages": 33,
+      "seconds": 1.8629912129999866,
+      "bytes": 1107242,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 2.737182930999552,
-      "bytes": 7527517,
-      "pages": 37,
+      "seconds": 1.2984945649986912,
+      "bytes": 7220956,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 0.9472420219999549,
-      "bytes": 626425,
-      "pages": 19,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3963090409997676,
-      "bytes": 10793699,
-      "pages": 45,
+      "seconds": 1.4053506099990045,
+      "bytes": 11055172,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.3661847930006843,
-      "bytes": 7242166,
+      "seconds": 0.8915826160009601,
+      "bytes": 2309383,
       "pages": 29,
+      "shards": 2
+    },
+    {
+      "seconds": 1.056229023000924,
+      "bytes": 11737562,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0256425439993109,
-      "bytes": 848800,
-      "pages": 26,
+      "seconds": 0.9255671579994669,
+      "bytes": 1181038,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.479707024000163,
-      "bytes": 13318632,
-      "pages": 43,
+      "seconds": 3.174646825998934,
+      "bytes": 11736199,
+      "pages": 33,
       "shards": 4
     },
     {
-      "seconds": 1.157918292999966,
-      "bytes": 10292312,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.0644383409999136,
+      "bytes": 5968582,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 2.16527700000006,
-      "bytes": 1181305,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.9677518210010021,
+      "bytes": 2711632,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.226145381000606,
-      "bytes": 7407097,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.13261040100042,
+      "bytes": 5584296,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 0.7928119780008274,
-      "bytes": 1141631,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.2312986870001623,
+      "bytes": 11373221,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.3008312719994137,
-      "bytes": 10073702,
-      "pages": 31,
+      "seconds": 1.0926308529997186,
+      "bytes": 7644433,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.3391879739992874,
-      "bytes": 13717263,
-      "pages": 44,
-      "shards": 4
+      "seconds": 1.0316966930004128,
+      "bytes": 5875792,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 0.882563871999082,
-      "bytes": 1071171,
+      "seconds": 0.9227722110008472,
+      "bytes": 1091891,
       "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.5854626939999434,
-      "bytes": 7158059,
-      "pages": 27,
-      "shards": 3
+      "seconds": 1.4008473989997583,
+      "bytes": 11311428,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1073119590000715,
-      "bytes": 10577372,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.1716579149997415,
+      "bytes": 6474393,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 0.7750392130001273,
-      "bytes": 1199165,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.9960888349996821,
+      "bytes": 2528980,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.0709359819993551,
-      "bytes": 7553738,
-      "pages": 32,
+      "seconds": 1.481450629999017,
+      "bytes": 11596666,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.9643511739996029,
-      "bytes": 6918272,
-      "pages": 35,
+      "seconds": 1.1495631180005148,
+      "bytes": 5725671,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.0286668139997346,
-      "bytes": 9530498,
-      "pages": 29,
+      "seconds": 1.1277136550015712,
+      "bytes": 7537147,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9772154700003739,
-      "bytes": 2553165,
-      "pages": 32,
+      "seconds": 1.212965857999734,
+      "bytes": 6163951,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3266675979994034,
-      "bytes": 5992849,
-      "pages": 30,
+      "seconds": 1.403984237998884,
+      "bytes": 849560,
+      "pages": 25,
+      "shards": 1
+    },
+    {
+      "seconds": 2.95492070399996,
+      "bytes": 11249015,
+      "pages": 38,
+      "shards": 4
+    },
+    {
+      "seconds": 1.3934125020005013,
+      "bytes": 6804482,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.2236139670003467,
-      "bytes": 7766768,
-      "pages": 37,
+      "seconds": 1.1164342709998891,
+      "bytes": 7715646,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.4731977419996838,
-      "bytes": 6241189,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.6021141640012502,
+      "bytes": 6045584,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.5922990050003136,
-      "bytes": 11693360,
-      "pages": 32,
+      "seconds": 1.4088935970012244,
+      "bytes": 4960887,
+      "pages": 38,
+      "shards": 2
+    },
+    {
+      "seconds": 2.058075558999917,
+      "bytes": 7841052,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8287107050000486,
-      "bytes": 2337614,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.1063804570003413,
+      "bytes": 836703,
+      "pages": 25,
+      "shards": 1
     },
     {
-      "seconds": 0.9556833940005163,
-      "bytes": 6215473,
+      "seconds": 1.6271591859986074,
+      "bytes": 16803522,
       "pages": 35,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.3753932430008717,
-      "bytes": 9934254,
-      "pages": 39,
+      "seconds": 1.121872658999564,
+      "bytes": 7611150,
+      "pages": 37,
       "shards": 3
     }
   ],

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4aadd2411e27d45ea0852b13570a9cc811de9a400c6b05705bce64c357639250
-size 721766463
+oid sha256:68b2bd34b599c12b63f5eb6b5156f7905dc9454684b6f321753aaa307c958f77
+size 721776127

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,