Qwen90M step 175000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 170000,
+  "step": 175000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "eae93b5a987696907c76cdf5c39f1ece9ef2dc6c0dacd54567e9b61a70909330",
+    "model.safetensors": "bf8bcea4b27ce5fc3908bc87b80c3e8677a65a70ce5cf629a31b7320189c02c4",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "68b2bd34b599c12b63f5eb6b5156f7905dc9454684b6f321753aaa307c958f77",
-    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
+    "trainer.pt": "561f426b2be02d82335b84193429431bda8e95fe9b5c193b863e61a80dca61f5",
+    "training_config.json": "a9be8fd9bf1c8d33aae0c56259fc899809df0d3ff60285ef0a4c2ea73471632c",
     "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
-    "shard_scheduler.json": "6b80239d2fad6471bd75c8996f07573047714dceff819ccf239934bc4693936c",
-    "coverage.json": "33a1ecc073a8506314c0c7e8b1455eaacfff9bce622ac5436c9d5ec2fcbfbf5d",
+    "shard_scheduler.json": "c42a0ac3d7e56da16d1f65cf4c4db0680516dfe3bc7cbf9a80b15627538de137",
+    "coverage.json": "4774588139ebf8c7826eaac403e76290c19e92b70a4de187efdb83579e13590f",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 170000,
+  "step": 175000,
   "stats": {
-    "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
+    "climbmix/blocks": 6300000,
+    "climbmix/targets": 10529447861,
+    "climbmix/input_tokens": 10530115542,
+    "ultra_qa/blocks": 2362500,
+    "ultra_qa/targets": 3854193552,
+    "ultra_qa/input_tokens": 3854200946,
+    "cortex/blocks": 2362500,
+    "cortex/targets": 89868402,
+    "cortex/input_tokens": 3755175304,
+    "ultra_style/blocks": 2362500,
+    "ultra_style/targets": 3736612861,
+    "ultra_style/input_tokens": 3736809089,
+    "openmath/blocks": 1575000,
+    "openmath/targets": 2433352240,
+    "openmath/input_tokens": 2434163081,
+    "rewrite6/blocks": 787500,
+    "rewrite6/targets": 1290978119,
+    "rewrite6/input_tokens": 1292393672
   },
   "sources": {
     "climbmix": {
-      "page": 838,
-      "row": 3535,
+      "page": 863,
+      "row": 1682,
       "start": {
-        "file": 40,
+        "file": 41,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 40,
+          "row_group": 58,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 3435983
+      "offset": 26624,
+      "documents": 3536530
     },
     "rewrite6": {
-      "page": 26,
-      "row": 4050,
+      "page": 41,
+      "row": 93,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 106496
+          "row": 167936
         }
       },
       "epoch": 9,
       "offset": 2048,
-      "documents": 1959182
+      "documents": 2016665
     },
     "openmath": {
-      "page": 523,
-      "row": 3399,
+      "page": 539,
+      "row": 1170,
       "start": {
-        "file": 4,
+        "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 395,
-          "row": 608
+          "row_group": 24,
+          "row": 494
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2145607
+      "offset": 0,
+      "documents": 2208914
     },
     "ultra_qa": {
-      "page": 226,
-      "row": 2070,
+      "page": 233,
+      "row": 645,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 18609
+          "row": 47281
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 927766
+      "offset": 0,
+      "documents": 955013
     },
     "ultra_style": {
-      "page": 345,
-      "row": 2329,
+      "page": 355,
+      "row": 3984,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 43238
+          "row": 84198
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1415449
+      "offset": 6144,
+      "documents": 1458064
     },
     "cortex": {
-      "page": 44505,
-      "row": 35,
+      "page": 45814,
+      "row": 100,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 5696675
+      "documents": 5864292
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"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"}
+{"step": 175000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "bf8bcea4b27ce5fc3908bc87b80c3e8677a65a70ce5cf629a31b7320189c02c4", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "561f426b2be02d82335b84193429431bda8e95fe9b5c193b863e61a80dca61f5", "training_config.json": "a9be8fd9bf1c8d33aae0c56259fc899809df0d3ff60285ef0a4c2ea73471632c", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "c42a0ac3d7e56da16d1f65cf4c4db0680516dfe3bc7cbf9a80b15627538de137", "coverage.json": "4774588139ebf8c7826eaac403e76290c19e92b70a4de187efdb83579e13590f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000175000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eae93b5a987696907c76cdf5c39f1ece9ef2dc6c0dacd54567e9b61a70909330
+oid sha256:bf8bcea4b27ce5fc3908bc87b80c3e8677a65a70ce5cf629a31b7320189c02c4
 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": 74872.19943677516,
-      "required_pages_per_second_upper_estimate": 0.0064897324853250855
+      "required_token_positions_per_second": 73516.90203215492,
+      "required_pages_per_second_upper_estimate": 0.006372258741262471
     },
     {
       "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": 9359.024929596895,
-      "required_pages_per_second_upper_estimate": 0.0037334549743086387
+      "required_token_positions_per_second": 9189.612754019365,
+      "required_pages_per_second_upper_estimate": 0.003665873924533016
     },
     {
       "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": 18718.04985919379,
-      "required_pages_per_second_upper_estimate": 0.004330091408498271
+      "required_token_positions_per_second": 18379.22550803873,
+      "required_pages_per_second_upper_estimate": 0.004251710357963462
     },
     {
       "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": 28077.074788790687,
-      "required_pages_per_second_upper_estimate": 0.0017097465456105776
+      "required_token_positions_per_second": 27568.838262058096,
+      "required_pages_per_second_upper_estimate": 0.0016787976076435398
     },
     {
       "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": 28077.074788790687,
-      "required_pages_per_second_upper_estimate": 0.002161450939558572
+      "required_token_positions_per_second": 27568.838262058096,
+      "required_pages_per_second_upper_estimate": 0.0021223254848419465
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 21,
-      "required_token_positions_per_second": 28077.074788790687,
-      "required_pages_per_second_upper_estimate": 0.34862330095224164
+      "required_token_positions_per_second": 27568.838262058096,
+      "required_pages_per_second_upper_estimate": 0.34231270424845844
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 61,
+    "prefetch_batches": 60,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "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,
+    "producer_batch_seconds": 0.45394105290606035,
+    "producer_p95_seconds": 0.30462472699946375,
+    "gpu_update_seconds": 1.0028714208843124,
+    "producer_batches_per_second": 2.2029291988423494,
+    "gpu_batches_per_second": 0.9971368005663374,
+    "producer_headroom": 2.2092547357505667,
     "sustainable": true,
-    "buffer_seconds": 60.06779597543113,
+    "buffer_seconds": 60.17228525305874,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 428682.5490002683,
-    "upload_bytes_per_second": 5225926.143782697,
+    "generation_bytes_per_second": 256138.59270394166,
+    "upload_bytes_per_second": 5381436.863729658,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 91.98529579645191,
+    "estimated_full_shards_per_hour": 54.96137939299285,
     "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": 1.0256425439993109,
-      "bytes": 848800,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.479707024000163,
-      "bytes": 13318632,
-      "pages": 43,
-      "shards": 4
+      "seconds": 1.13261040100042,
+      "bytes": 5584296,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.157918292999966,
-      "bytes": 10292312,
+      "seconds": 1.2312986870001623,
+      "bytes": 11373221,
       "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 2.16527700000006,
-      "bytes": 1181305,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.0926308529997186,
+      "bytes": 7644433,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.226145381000606,
-      "bytes": 7407097,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.0316966930004128,
+      "bytes": 5875792,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 0.7928119780008274,
-      "bytes": 1141631,
+      "seconds": 0.9227722110008472,
+      "bytes": 1091891,
       "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.3008312719994137,
-      "bytes": 10073702,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3391879739992874,
-      "bytes": 13717263,
-      "pages": 44,
+      "seconds": 1.4008473989997583,
+      "bytes": 11311428,
+      "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 0.882563871999082,
-      "bytes": 1071171,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1716579149997415,
+      "bytes": 6474393,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.5854626939999434,
-      "bytes": 7158059,
-      "pages": 27,
+      "seconds": 0.9960888349996821,
+      "bytes": 2528980,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.1073119590000715,
-      "bytes": 10577372,
-      "pages": 37,
+      "seconds": 1.481450629999017,
+      "bytes": 11596666,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.7750392130001273,
-      "bytes": 1199165,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.1495631180005148,
+      "bytes": 5725671,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.0709359819993551,
-      "bytes": 7553738,
-      "pages": 32,
+      "seconds": 1.1277136550015712,
+      "bytes": 7537147,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9643511739996029,
-      "bytes": 6918272,
-      "pages": 35,
+      "seconds": 1.212965857999734,
+      "bytes": 6163951,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0286668139997346,
-      "bytes": 9530498,
-      "pages": 29,
-      "shards": 3
+      "seconds": 1.403984237998884,
+      "bytes": 849560,
+      "pages": 25,
+      "shards": 1
     },
     {
-      "seconds": 0.9772154700003739,
-      "bytes": 2553165,
-      "pages": 32,
-      "shards": 2
+      "seconds": 2.95492070399996,
+      "bytes": 11249015,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.3266675979994034,
-      "bytes": 5992849,
-      "pages": 30,
+      "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
     },
     {
-      "seconds": 0.9314350770000601,
-      "bytes": 2320145,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.0686068680006429,
+      "bytes": 1192521,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.1146210560000327,
-      "bytes": 5994268,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.109725468999386,
+      "bytes": 7035828,
+      "pages": 21,
+      "shards": 3
     },
     {
-      "seconds": 0.7972013050002715,
-      "bytes": 688709,
-      "pages": 19,
+      "seconds": 0.8532845239988092,
+      "bytes": 842129,
+      "pages": 25,
       "shards": 1
     },
     {
-      "seconds": 1.845211140000174,
-      "bytes": 18359200,
-      "pages": 44,
-      "shards": 5
-    },
-    {
-      "seconds": 0.8465310680003313,
-      "bytes": 1116330,
-      "pages": 33,
-      "shards": 1
+      "seconds": 0.9992884630009939,
+      "bytes": 9868032,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.1592603440003586,
-      "bytes": 7563694,
-      "pages": 38,
+      "seconds": 1.1347405200012872,
+      "bytes": 8295093,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1315413200009061,
-      "bytes": 5996697,
+      "seconds": 1.3122159340000508,
+      "bytes": 6195105,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 2.1342484679989866,
-      "bytes": 1010482,
-      "pages": 32,
-      "shards": 1
+      "seconds": 1.4069043029994646,
+      "bytes": 6338726,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.0687636729999213,
-      "bytes": 11353142,
-      "pages": 26,
-      "shards": 4
+      "seconds": 1.157190105999689,
+      "bytes": 7618038,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.1774445949995425,
-      "bytes": 12065836,
-      "pages": 43,
-      "shards": 3
+      "seconds": 1.2278081520016713,
+      "bytes": 6384943,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.8629912129999866,
-      "bytes": 1107242,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.068345749998116,
+      "bytes": 2639565,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.2984945649986912,
-      "bytes": 7220956,
-      "pages": 28,
-      "shards": 3
+      "seconds": 1.2386211259981792,
+      "bytes": 5629626,
+      "pages": 20,
+      "shards": 2
     },
     {
-      "seconds": 1.4053506099990045,
-      "bytes": 11055172,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.9895669670004281,
+      "bytes": 15602081,
+      "pages": 47,
+      "shards": 4
     },
     {
-      "seconds": 0.8915826160009601,
-      "bytes": 2309383,
-      "pages": 29,
+      "seconds": 0.9570206669995969,
+      "bytes": 2289232,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.056229023000924,
-      "bytes": 11737562,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.238607878000039,
+      "bytes": 6177061,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9255671579994669,
-      "bytes": 1181038,
-      "pages": 34,
+      "seconds": 1.0433953600004315,
+      "bytes": 1237044,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 3.174646825998934,
-      "bytes": 11736199,
-      "pages": 33,
+      "seconds": 1.392788182998629,
+      "bytes": 11322146,
+      "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 1.0644383409999136,
-      "bytes": 5968582,
-      "pages": 35,
+      "seconds": 1.1545408300007693,
+      "bytes": 6140094,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.9677518210010021,
-      "bytes": 2711632,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1522124919974885,
+      "bytes": 8515706,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.13261040100042,
-      "bytes": 5584296,
-      "pages": 27,
+      "seconds": 1.1056661020011234,
+      "bytes": 6106174,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2312986870001623,
-      "bytes": 11373221,
+      "seconds": 1.0782650920009473,
+      "bytes": 5102757,
       "pages": 36,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.0926308529997186,
-      "bytes": 7644433,
-      "pages": 38,
+      "seconds": 1.0556685640003707,
+      "bytes": 7468948,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.0316966930004128,
-      "bytes": 5875792,
-      "pages": 27,
+      "seconds": 1.1485331379990384,
+      "bytes": 5965663,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.9227722110008472,
-      "bytes": 1091891,
-      "pages": 33,
+      "seconds": 0.951670551999996,
+      "bytes": 914174,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.4008473989997583,
-      "bytes": 11311428,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.4923641690002114,
+      "bytes": 17877863,
+      "pages": 40,
+      "shards": 5
     },
     {
-      "seconds": 1.1716579149997415,
-      "bytes": 6474393,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.7990716060012346,
+      "bytes": 1163166,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 0.9960888349996821,
-      "bytes": 2528980,
-      "pages": 33,
+      "seconds": 1.5562365400001,
+      "bytes": 7212570,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.481450629999017,
-      "bytes": 11596666,
-      "pages": 35,
-      "shards": 3
+      "seconds": 0.9564802560016687,
+      "bytes": 6117579,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.1495631180005148,
-      "bytes": 5725671,
-      "pages": 33,
+      "seconds": 1.0467319720009982,
+      "bytes": 5732511,
+      "pages": 41,
       "shards": 2
     },
     {
-      "seconds": 1.1277136550015712,
-      "bytes": 7537147,
-      "pages": 37,
+      "seconds": 1.2224297300017497,
+      "bytes": 7561441,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.212965857999734,
-      "bytes": 6163951,
-      "pages": 36,
+      "seconds": 1.1380056220004917,
+      "bytes": 6964732,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.403984237998884,
-      "bytes": 849560,
-      "pages": 25,
-      "shards": 1
+      "seconds": 1.278426272001525,
+      "bytes": 5758474,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 2.95492070399996,
-      "bytes": 11249015,
-      "pages": 38,
+      "seconds": 1.1922402710006281,
+      "bytes": 10988746,
+      "pages": 34,
       "shards": 4
     },
     {
-      "seconds": 1.3934125020005013,
-      "bytes": 6804482,
+      "seconds": 0.8872451259994705,
+      "bytes": 1248151,
       "pages": 34,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.1164342709998891,
-      "bytes": 7715646,
-      "pages": 35,
+      "seconds": 1.248066406998987,
+      "bytes": 7262597,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.6021141640012502,
-      "bytes": 6045584,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.8794477230003395,
+      "bytes": 663465,
+      "pages": 19,
+      "shards": 1
     },
     {
-      "seconds": 1.4088935970012244,
-      "bytes": 4960887,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.393969371001731,
+      "bytes": 16331456,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 2.058075558999917,
-      "bytes": 7841052,
-      "pages": 38,
+      "seconds": 1.0225603920007416,
+      "bytes": 7736202,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1063804570003413,
-      "bytes": 836703,
-      "pages": 25,
+      "seconds": 1.0140705839985458,
+      "bytes": 1147707,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.6271591859986074,
-      "bytes": 16803522,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.227832770000532,
+      "bytes": 5733304,
+      "pages": 21,
+      "shards": 2
     },
     {
-      "seconds": 1.121872658999564,
-      "bytes": 7611150,
-      "pages": 37,
+      "seconds": 1.3411155760004476,
+      "bytes": 6738025,
+      "pages": 38,
       "shards": 3
+    },
+    {
+      "seconds": 1.1425984330016945,
+      "bytes": 6317215,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 0.939828511000087,
+      "bytes": 6670718,
+      "pages": 28,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68b2bd34b599c12b63f5eb6b5156f7905dc9454684b6f321753aaa307c958f77
-size 721776127
+oid sha256:561f426b2be02d82335b84193429431bda8e95fe9b5c193b863e61a80dca61f5
+size 721759231

training_config.json

@@ -46,7 +46,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 61,
+  "prefetch_batches": 60,
   "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,