Qwen90M step 735000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 730000,
+  "step": 735000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "8c53eaad59485c3e7f93abf3a4b58bce24607b53a618b83f0a58c30a6ceb0a27",
+    "model.safetensors": "a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "1650174bff312eebb6355ae334a6607b73e2c555e90fe45a7a399e9567721584",
-    "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687",
+    "trainer.pt": "3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315",
+    "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037",
     "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379",
-    "shard_scheduler.json": "1ca6461804592e6d06e70f93e961d9b78fbee44b41acd3825f26f6670f746dd6",
-    "coverage.json": "a190366ff62a7a816e8e5cfff69e551cc2bb0b47d066c9abcec59ff0915c60cf",
+    "shard_scheduler.json": "541a1f3f705d2b085cbb51195f8b00accd55d018941e78dc4445045fc4c03dcd",
+    "coverage.json": "fe7d1dc256a120fa7e42273b21198470f2acfeaaf06e57edd8a92aeb71326f7c",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 730000,
+  "step": 735000,
   "stats": {
-    "climbmix/blocks": 26280000,
-    "climbmix/targets": 43931838219,
-    "climbmix/input_tokens": 43934622432,
-    "ultra_qa/blocks": 9855000,
-    "ultra_qa/targets": 16077362872,
-    "ultra_qa/input_tokens": 16077393915,
-    "cortex/blocks": 9855000,
-    "cortex/targets": 374703742,
-    "cortex/input_tokens": 15664993820,
-    "ultra_style/blocks": 9855000,
-    "ultra_style/targets": 15587572089,
-    "ultra_style/input_tokens": 15588389892,
-    "openmath/blocks": 6570000,
-    "openmath/targets": 10152051955,
-    "openmath/input_tokens": 10155434511,
-    "rewrite6/blocks": 3285000,
-    "rewrite6/targets": 5385363851,
-    "rewrite6/input_tokens": 5391268095
+    "climbmix/blocks": 26460000,
+    "climbmix/targets": 44233581693,
+    "climbmix/input_tokens": 44236384555,
+    "ultra_qa/blocks": 9922500,
+    "ultra_qa/targets": 16187472584,
+    "ultra_qa/input_tokens": 16187503812,
+    "cortex/blocks": 9922500,
+    "cortex/targets": 377265616,
+    "cortex/input_tokens": 15772193657,
+    "ultra_style/blocks": 9922500,
+    "ultra_style/targets": 15694417562,
+    "ultra_style/input_tokens": 15695240794,
+    "openmath/blocks": 6615000,
+    "openmath/targets": 10221476545,
+    "openmath/input_tokens": 10224882264,
+    "rewrite6/blocks": 3307500,
+    "rewrite6/targets": 5422206078,
+    "rewrite6/input_tokens": 5428150619
   },
   "sources": {
     "climbmix": {
-      "page": 3599,
-      "row": 1241,
+      "page": 3623,
+      "row": 2520,
       "start": {
-        "file": 174,
+        "file": 175,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 16,
+          "row_group": 31,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 14742745
+      "offset": 2048,
+      "documents": 14842328
     },
     "rewrite6": {
-      "page": 47,
-      "row": 2791,
+      "page": 11,
+      "row": 2259,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 192512
+          "row": 45056
         }
       },
-      "epoch": 40,
-      "offset": 2048,
-      "documents": 8411463
+      "epoch": 41,
+      "offset": 0,
+      "documents": 8468879
     },
     "openmath": {
-      "page": 2249,
-      "row": 1624,
+      "page": 2264,
+      "row": 3335,
       "start": {
         "file": 21,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 42,
-          "row": 254
+          "row_group": 103,
+          "row": 694
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 9213528
+      "documents": 9276679
     },
     "ultra_qa": {
-      "page": 972,
-      "row": 2405,
+      "page": 979,
+      "row": 1035,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 21207
+          "row": 49879
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 3983717
+      "offset": 4096,
+      "documents": 4011019
     },
     "ultra_style": {
-      "page": 1484,
-      "row": 2035,
+      "page": 1494,
+      "row": 2372,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 2817
+          "row": 43777
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 6080499
+      "offset": 2048,
+      "documents": 6121796
     },
     "cortex": {
-      "page": 191125,
-      "row": 60,
+      "page": 192433,
+      "row": 128,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 24464060
+      "documents": 24631552
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 730000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "8c53eaad59485c3e7f93abf3a4b58bce24607b53a618b83f0a58c30a6ceb0a27", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1650174bff312eebb6355ae334a6607b73e2c555e90fe45a7a399e9567721584", "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "1ca6461804592e6d06e70f93e961d9b78fbee44b41acd3825f26f6670f746dd6", "coverage.json": "a190366ff62a7a816e8e5cfff69e551cc2bb0b47d066c9abcec59ff0915c60cf", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000730000"}
+{"step": 735000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315", "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "541a1f3f705d2b085cbb51195f8b00accd55d018941e78dc4445045fc4c03dcd", "coverage.json": "fe7d1dc256a120fa7e42273b21198470f2acfeaaf06e57edd8a92aeb71326f7c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000735000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c53eaad59485c3e7f93abf3a4b58bce24607b53a618b83f0a58c30a6ceb0a27
+oid sha256:a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73
 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": 76313.75839753207,
-      "required_pages_per_second_upper_estimate": 0.005279562878256555
+      "required_token_positions_per_second": 76078.75752915804,
+      "required_pages_per_second_upper_estimate": 0.0052633049729053005
     },
     {
       "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": 9539.21979969151,
-      "required_pages_per_second_upper_estimate": 0.003993872128569453
+      "required_token_positions_per_second": 9509.844691144755,
+      "required_pages_per_second_upper_estimate": 0.003981573384042948
     },
     {
       "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": 19078.43959938302,
-      "required_pages_per_second_upper_estimate": 0.003973535836265066
+      "required_token_positions_per_second": 19019.68938228951,
+      "required_pages_per_second_upper_estimate": 0.0039612997153919026
     },
     {
       "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": 28617.659399074528,
-      "required_pages_per_second_upper_estimate": 0.0017155849843387327
+      "required_token_positions_per_second": 28529.534073434268,
+      "required_pages_per_second_upper_estimate": 0.0017103020056261802
     },
     {
       "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": 28617.659399074528,
-      "required_pages_per_second_upper_estimate": 0.002224897142497311
+      "required_token_positions_per_second": 28529.534073434268,
+      "required_pages_per_second_upper_estimate": 0.0022180457860511227
     },
     {
       "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": 28617.659399074528,
-      "required_pages_per_second_upper_estimate": 0.3670766075226655
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28529.534073434268,
+      "required_pages_per_second_upper_estimate": 0.36594623046695485
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,180 +155,30 @@
   "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": 0.10201518258595854,
-    "producer_p95_seconds": 0.18670680400100537,
-    "gpu_update_seconds": 0.9661167468117297,
-    "producer_batches_per_second": 9.802462483046527,
-    "gpu_batches_per_second": 1.035071592848471,
-    "producer_headroom": 9.47032316486494,
+    "producer_batch_seconds": 0.08382008823436138,
+    "producer_p95_seconds": 0.23780670700034534,
+    "gpu_update_seconds": 0.9691009999965221,
+    "producer_batches_per_second": 11.930314332335168,
+    "gpu_batches_per_second": 1.031884189577339,
+    "producer_headroom": 11.561679549738852,
     "sustainable": true,
-    "buffer_seconds": 60.86535504913897,
+    "buffer_seconds": 60.084261999784374,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 342045.38893318584,
-    "upload_bytes_per_second": 4834421.459987242,
+    "generation_bytes_per_second": 304411.78655841085,
+    "upload_bytes_per_second": 4856659.550780564,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 73.39497805592234,
+    "estimated_full_shards_per_hour": 65.319683051722,
     "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": 2.1827129590001277,
-      "bytes": 47377,
-      "pages": 1,
-      "shards": 1
-    },
-    {
-      "seconds": 1.064991151999493,
-      "bytes": 1003408,
-      "pages": 31,
-      "shards": 1
-    },
-    {
-      "seconds": 1.010816134000379,
-      "bytes": 5780671,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.7082208169995283,
-      "bytes": 12327681,
-      "pages": 32,
-      "shards": 4
-    },
-    {
-      "seconds": 2.076565071999539,
-      "bytes": 6113882,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1128287640003691,
-      "bytes": 6043549,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0340174030006892,
-      "bytes": 2678432,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1785862379992977,
-      "bytes": 9722174,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4900774309999179,
-      "bytes": 2616245,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2741041360004601,
-      "bytes": 11482818,
-      "pages": 21,
-      "shards": 3
-    },
-    {
-      "seconds": 1.076980086999356,
-      "bytes": 1228924,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3698066919996563,
-      "bytes": 11420874,
-      "pages": 33,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0434529929998462,
-      "bytes": 6181688,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.099565554000037,
-      "bytes": 1093470,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3861621640007797,
-      "bytes": 7683491,
-      "pages": 35,
-      "shards": 4
-    },
-    {
-      "seconds": 1.2777914989992496,
-      "bytes": 16450745,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9403941730006409,
-      "bytes": 2445089,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.042657115000111,
-      "bytes": 6054602,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0668450510002003,
-      "bytes": 1106491,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.397956458000408,
-      "bytes": 11076576,
-      "pages": 34,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1856196190001356,
-      "bytes": 6180989,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3201266620008028,
-      "bytes": 6761235,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2538779149999755,
-      "bytes": 7546399,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2998433860002478,
-      "bytes": 5962432,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2994795300000987,
-      "bytes": 7606454,
-      "pages": 39,
-      "shards": 3
-    },
     {
       "seconds": 1.54305899999963,
       "bytes": 5745307,
@@ -466,6 +316,252 @@
       "bytes": 10315993,
       "pages": 33,
       "shards": 3
+    },
+    {
+      "seconds": 1.4511640289983916,
+      "bytes": 7595757,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2930020470012096,
+      "bytes": 547633,
+      "pages": 17,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3445512510006665,
+      "bytes": 5828365,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2113412039998366,
+      "bytes": 6839007,
+      "pages": 41,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4825698250006099,
+      "bytes": 11644292,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3249330049984565,
+      "bytes": 7496918,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1885178210013692,
+      "bytes": 1081987,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.9165967070002807,
+      "bytes": 10316454,
+      "pages": 33,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1245960120013478,
+      "bytes": 2666345,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2064912770001683,
+      "bytes": 6004944,
+      "pages": 31,
+      "shards": 2
+    },
+    {
+      "seconds": 3.7895365029999084,
+      "bytes": 13350183,
+      "pages": 36,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1515760519996547,
+      "bytes": 5491935,
+      "pages": 30,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1848105680001026,
+      "bytes": 6290344,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.0463193560008222,
+      "bytes": 2621559,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.171350075999726,
+      "bytes": 6139992,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.280977876000179,
+      "bytes": 9975922,
+      "pages": 33,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2528174399994896,
+      "bytes": 8102539,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1965843840007437,
+      "bytes": 6380308,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2874604959997669,
+      "bytes": 2530940,
+      "pages": 33,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3693337599997903,
+      "bytes": 10798536,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.092897812000956,
+      "bytes": 6537422,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9582660920004855,
+      "bytes": 2603502,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.5727497110001423,
+      "bytes": 11757823,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1941701559990179,
+      "bytes": 5221736,
+      "pages": 34,
+      "shards": 2
+    },
+    {
+      "seconds": 1.6056268569991516,
+      "bytes": 7578967,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.20693444599965,
+      "bytes": 6453330,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9651811909989192,
+      "bytes": 2682508,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4213833419998991,
+      "bytes": 10754179,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.5129684479998105,
+      "bytes": 6722253,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.165229547999843,
+      "bytes": 7984438,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.001051560999258,
+      "bytes": 939443,
+      "pages": 30,
+      "shards": 1
+    },
+    {
+      "seconds": 1.4753997340012575,
+      "bytes": 11016252,
+      "pages": 34,
+      "shards": 4
+    },
+    {
+      "seconds": 1.6974678700007644,
+      "bytes": 6266676,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.6492898739998054,
+      "bytes": 1094803,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.8326161909990333,
+      "bytes": 13393499,
+      "pages": 34,
+      "shards": 4
+    },
+    {
+      "seconds": 1.4041516550005326,
+      "bytes": 11034703,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1081106500005262,
+      "bytes": 1056276,
+      "pages": 28,
+      "shards": 1
+    },
+    {
+      "seconds": 1.211192590999417,
+      "bytes": 7370118,
+      "pages": 35,
+      "shards": 3
+    },
+    {
+      "seconds": 1.246054737999657,
+      "bytes": 1155172,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.6413428979994933,
+      "bytes": 11205821,
+      "pages": 33,
+      "shards": 4
+    },
+    {
+      "seconds": 1.3393069079993438,
+      "bytes": 11931034,
+      "pages": 38,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1650174bff312eebb6355ae334a6607b73e2c555e90fe45a7a399e9567721584
-size 721792959
+oid sha256:3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315
+size 721775167

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,