Qwen90M step 915000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 910000,
+  "step": 915000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "b4df0ef53a5a343fee1e4dce6ab8a731c5f01c993a900224ba917205158c612a",
+    "model.safetensors": "aeb6dc7998f5f2de45ed9d74ce8a305219296bca84ed4d28eb9e700409a7e925",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "8ba9bbe2eb54c264d2b9324ec47dc07ced3c51c8d254214772fea4aa3594fd98",
-    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
+    "trainer.pt": "b5920c2beb588bacfe2a156d706d2a8ce72852d4602bd521d016d792453dc504",
+    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "dd11a2928b6fc6dbcfc1eafd5d86aa23cf716d5b4273f83412bf2508c81e3964",
-    "coverage.json": "26aa8059e8f72378a7c5a4bd2300bf3c7cff681ad595ec96d08d6b092bc45174",
+    "shard_scheduler.json": "6c7fa1608e35806ad39bf2768b3b42afcb2a61f8885c179d6a26d6ab3767bb17",
+    "coverage.json": "f9a9a9cfc3fffe4de28b077d555cd2d486c5f650819afb6fae351ad9c128f3bc",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 910000,
+  "step": 915000,
   "stats": {
-    "climbmix/blocks": 32760000,
-    "climbmix/targets": 54764250124,
-    "climbmix/input_tokens": 54767720896,
-    "ultra_qa/blocks": 12285000,
-    "ultra_qa/targets": 20042043648,
-    "ultra_qa/input_tokens": 20042082120,
-    "cortex/blocks": 12285000,
-    "cortex/targets": 467130820,
-    "cortex/input_tokens": 19527965675,
-    "ultra_style/blocks": 12285000,
-    "ultra_style/targets": 19430317419,
-    "ultra_style/input_tokens": 19431336007,
-    "openmath/blocks": 8190000,
-    "openmath/targets": 12654930072,
-    "openmath/input_tokens": 12659146541,
-    "rewrite6/blocks": 4095000,
-    "rewrite6/targets": 6713205114,
-    "rewrite6/input_tokens": 6720565027
+    "climbmix/blocks": 32940000,
+    "climbmix/targets": 55064840955,
+    "climbmix/input_tokens": 55068330846,
+    "ultra_qa/blocks": 12352500,
+    "ultra_qa/targets": 20152216028,
+    "ultra_qa/input_tokens": 20152254712,
+    "cortex/blocks": 12352500,
+    "cortex/targets": 469703401,
+    "cortex/input_tokens": 19635241143,
+    "ultra_style/blocks": 12352500,
+    "ultra_style/targets": 19536930397,
+    "ultra_style/input_tokens": 19537954240,
+    "openmath/blocks": 8235000,
+    "openmath/targets": 12724599700,
+    "openmath/input_tokens": 12728839417,
+    "rewrite6/blocks": 4117500,
+    "rewrite6/targets": 6750063827,
+    "rewrite6/input_tokens": 6757464407
   },
   "sources": {
     "climbmix": {
-      "page": 4486,
-      "row": 2973,
+      "page": 4511,
+      "row": 1767,
       "start": {
-        "file": 217,
+        "file": 218,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 13,
+          "row_group": 30,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 18377629
+      "offset": 2048,
+      "documents": 18478823
     },
     "rewrite6": {
-      "page": 2,
-      "row": 1549,
+      "page": 16,
+      "row": 2035,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 8192
+          "row": 65536
         }
       },
       "epoch": 51,
-      "offset": 2048,
-      "documents": 10485345
+      "offset": 0,
+      "documents": 10543175
     },
     "openmath": {
-      "page": 2804,
-      "row": 458,
+      "page": 2819,
+      "row": 2187,
       "start": {
         "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 132,
-          "row": 287
+          "row_group": 193,
+          "row": 727
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11485642
+      "documents": 11548811
     },
     "ultra_qa": {
-      "page": 1212,
-      "row": 1525,
+      "page": 1219,
+      "row": 126,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 29416
+          "row": 58088
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4965877
+      "documents": 4993150
     },
     "ultra_style": {
-      "page": 1850,
-      "row": 2060,
+      "page": 1860,
+      "row": 2111,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 43244
+          "row": 84204
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 7579660
+      "offset": 2048,
+      "documents": 7620671
     },
     "cortex": {
-      "page": 238252,
-      "row": 11,
+      "page": 239561,
+      "row": 27,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 30496267
+      "documents": 30663835
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 910000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b4df0ef53a5a343fee1e4dce6ab8a731c5f01c993a900224ba917205158c612a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8ba9bbe2eb54c264d2b9324ec47dc07ced3c51c8d254214772fea4aa3594fd98", "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "dd11a2928b6fc6dbcfc1eafd5d86aa23cf716d5b4273f83412bf2508c81e3964", "coverage.json": "26aa8059e8f72378a7c5a4bd2300bf3c7cff681ad595ec96d08d6b092bc45174", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000910000"}
+{"step": 915000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "aeb6dc7998f5f2de45ed9d74ce8a305219296bca84ed4d28eb9e700409a7e925", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "b5920c2beb588bacfe2a156d706d2a8ce72852d4602bd521d016d792453dc504", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "6c7fa1608e35806ad39bf2768b3b42afcb2a61f8885c179d6a26d6ab3767bb17", "coverage.json": "f9a9a9cfc3fffe4de28b077d555cd2d486c5f650819afb6fae351ad9c128f3bc", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000915000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4df0ef53a5a343fee1e4dce6ab8a731c5f01c993a900224ba917205158c612a
+oid sha256:aeb6dc7998f5f2de45ed9d74ce8a305219296bca84ed4d28eb9e700409a7e925
 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": 76105.76941737354,
-      "required_pages_per_second_upper_estimate": 0.0052651737180082646
+      "required_token_positions_per_second": 76751.17991915937,
+      "required_pages_per_second_upper_estimate": 0.0053098247140804964
     },
     {
       "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": 9513.221177171692,
-      "required_pages_per_second_upper_estimate": 0.00398298704823338
+      "required_token_positions_per_second": 9593.897489894922,
+      "required_pages_per_second_upper_estimate": 0.004016764535657612
     },
     {
       "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": 19026.442354343384,
-      "required_pages_per_second_upper_estimate": 0.003962706181382876
+      "required_token_positions_per_second": 19187.794979789844,
+      "required_pages_per_second_upper_estimate": 0.003996311678108493
     },
     {
       "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": 28539.663531515074,
-      "required_pages_per_second_upper_estimate": 0.0017109092511713377
+      "required_token_positions_per_second": 28781.692469684764,
+      "required_pages_per_second_upper_estimate": 0.00172541851645782
     },
     {
       "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": 28539.663531515074,
-      "required_pages_per_second_upper_estimate": 0.002218833306862128
+      "required_token_positions_per_second": 28781.692469684764,
+      "required_pages_per_second_upper_estimate": 0.00223764999223203
     },
     {
       "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": 22,
-      "required_token_positions_per_second": 28539.663531515074,
-      "required_pages_per_second_upper_estimate": 0.3660761602790507
+      "estimated_pages_for_buffer": 23,
+      "required_token_positions_per_second": 28781.692469684764,
+      "required_pages_per_second_upper_estimate": 0.36918064762746455
     }
   ],
   "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": 62,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.1008662234296196,
-    "producer_p95_seconds": 0.2910494249990734,
-    "gpu_update_seconds": 0.9687570412128212,
-    "producer_batches_per_second": 9.914121556238891,
-    "gpu_batches_per_second": 1.0322505617590811,
-    "producer_headroom": 9.60437506504624,
+    "producer_batch_seconds": 0.09606207005469969,
+    "producer_p95_seconds": 0.27300766600092174,
+    "gpu_update_seconds": 0.9606106391804838,
+    "producer_batches_per_second": 10.409935986498935,
+    "gpu_batches_per_second": 1.0410045019417233,
+    "producer_headroom": 9.999895261818663,
     "sustainable": true,
-    "buffer_seconds": 60.062936555194916,
+    "buffer_seconds": 60.51847026837048,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 371269.35917070275,
-    "upload_bytes_per_second": 4902520.230367588,
+    "generation_bytes_per_second": 235141.59924877153,
+    "upload_bytes_per_second": 4647383.918342561,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 79.66576176968395,
+    "estimated_full_shards_per_hour": 50.45591338250503,
     "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": 1.3777095789992018,
-      "bytes": 11141641,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 3.434384232998127,
-      "bytes": 5971898,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1304158860002644,
-      "bytes": 8230136,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2011196589992323,
-      "bytes": 6075797,
+      "seconds": 1.6493652240023948,
+      "bytes": 6440305,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.276424440999108,
-      "bytes": 6070440,
-      "pages": 25,
-      "shards": 3
+      "seconds": 1.0658796659990912,
+      "bytes": 2378111,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.2386367050021363,
-      "bytes": 6009209,
+      "seconds": 1.2226023729999724,
+      "bytes": 6024557,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1234037470021576,
-      "bytes": 6172938,
+      "seconds": 2.320106137998664,
+      "bytes": 11400278,
       "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8850959770024929,
-      "bytes": 2438669,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.6140179690009973,
-      "bytes": 15868305,
-      "pages": 31,
       "shards": 4
     },
     {
-      "seconds": 1.348259312002483,
-      "bytes": 5928296,
-      "pages": 35,
+      "seconds": 1.142467518999183,
+      "bytes": 6818785,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 0.9825653119987692,
-      "bytes": 2330980,
-      "pages": 28,
+      "seconds": 1.1884660100004112,
+      "bytes": 6414581,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2888210940000135,
-      "bytes": 6181272,
-      "pages": 37,
+      "seconds": 0.931925422999484,
+      "bytes": 2347633,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1400288160002674,
-      "bytes": 6499509,
-      "pages": 34,
+      "seconds": 1.7378346169971337,
+      "bytes": 10782839,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.1078390350012342,
-      "bytes": 6180333,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.3063333719983348,
+      "bytes": 7915290,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1313278150009864,
-      "bytes": 6410101,
-      "pages": 20,
-      "shards": 2
+      "seconds": 1.0522420070010412,
+      "bytes": 1136593,
+      "pages": 36,
+      "shards": 1
     },
     {
-      "seconds": 1.2548322069997084,
-      "bytes": 7827361,
-      "pages": 41,
+      "seconds": 1.2345661709987326,
+      "bytes": 11661576,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.3341312570009904,
-      "bytes": 10704029,
+      "seconds": 1.2733937859993603,
+      "bytes": 6517566,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.136130429997138,
-      "bytes": 965107,
-      "pages": 29,
-      "shards": 1
+      "seconds": 1.2599434170006134,
+      "bytes": 6272622,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.3451466709993838,
-      "bytes": 7707115,
+      "seconds": 1.9948137850005878,
+      "bytes": 7735186,
       "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 2.3437790619973384,
-      "bytes": 11207805,
-      "pages": 40,
-      "shards": 3
+      "seconds": 1.6032572549993347,
+      "bytes": 5165320,
+      "pages": 25,
+      "shards": 2
     },
     {
-      "seconds": 2.2792270009995264,
-      "bytes": 8233418,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.8941422179996152,
+      "bytes": 6019433,
+      "pages": 39,
+      "shards": 2
     },
     {
-      "seconds": 1.1861695589977899,
-      "bytes": 6023568,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.3825668060017051,
+      "bytes": 13491032,
+      "pages": 40,
+      "shards": 5
     },
     {
-      "seconds": 1.0451825370000734,
-      "bytes": 965218,
-      "pages": 28,
+      "seconds": 0.9105340399983106,
+      "bytes": 1249369,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.2396269109995046,
-      "bytes": 7721145,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.295263830001204,
+      "bytes": 5736433,
+      "pages": 23,
+      "shards": 2
     },
     {
-      "seconds": 1.3206068420004158,
-      "bytes": 11244076,
-      "pages": 40,
+      "seconds": 1.218638172998908,
+      "bytes": 7112834,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 1.1782944210026471,
-      "bytes": 2347277,
-      "pages": 29,
+      "seconds": 1.2346669860016846,
+      "bytes": 6391793,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2341140790013014,
-      "bytes": 11781395,
-      "pages": 31,
-      "shards": 4
+      "seconds": 1.537398492997454,
+      "bytes": 7291338,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 2.1994080930016935,
-      "bytes": 6022375,
+      "seconds": 1.5484525030005898,
+      "bytes": 10816659,
       "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5225848519985448,
-      "bytes": 6009508,
-      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.197222058999614,
-      "bytes": 6186205,
-      "pages": 37,
+      "seconds": 2.108060000999103,
+      "bytes": 5732876,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.0562214180026785,
-      "bytes": 2441509,
-      "pages": 32,
+      "seconds": 1.102688891998696,
+      "bytes": 2347731,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3156736879973323,
-      "bytes": 6047068,
-      "pages": 34,
+      "seconds": 1.3670132119987102,
+      "bytes": 6121191,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.5381603519999771,
-      "bytes": 16328441,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.156715154000267,
+      "bytes": 10313432,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.2096843409999565,
-      "bytes": 2710723,
+      "seconds": 1.0891685620008502,
+      "bytes": 2577323,
       "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.2958958110029926,
-      "bytes": 5955386,
-      "pages": 27,
-      "shards": 2
-    },
-    {
-      "seconds": 1.584322863000125,
-      "bytes": 6087710,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.423857304998819,
+      "bytes": 12061949,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.2062028339969402,
-      "bytes": 5756694,
-      "pages": 35,
+      "seconds": 1.17036508000092,
+      "bytes": 7649153,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.087463227002445,
-      "bytes": 6477149,
-      "pages": 36,
+      "seconds": 1.2898340729989286,
+      "bytes": 5540548,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.7014945069968235,
-      "bytes": 13448331,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.2250885649991687,
+      "bytes": 5931268,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.4291834129980998,
-      "bytes": 4870018,
-      "pages": 31,
+      "seconds": 1.6209707380003238,
+      "bytes": 2622297,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.6493652240023948,
-      "bytes": 6440305,
+      "seconds": 1.044743833001121,
+      "bytes": 6332578,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0658796659990912,
-      "bytes": 2378111,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.5906551469997794,
+      "bytes": 16693261,
+      "pages": 23,
+      "shards": 5
     },
     {
-      "seconds": 1.2226023729999724,
-      "bytes": 6024557,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.596418191002158,
+      "bytes": 1389732,
+      "pages": 43,
+      "shards": 1
     },
     {
-      "seconds": 2.320106137998664,
-      "bytes": 11400278,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.2193843360000756,
+      "bytes": 5952773,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.142467518999183,
-      "bytes": 6818785,
+      "seconds": 1.8950818810008059,
+      "bytes": 7230636,
       "pages": 33,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.1884660100004112,
-      "bytes": 6414581,
+      "seconds": 1.5622936799991294,
+      "bytes": 5942154,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.931925422999484,
-      "bytes": 2347633,
-      "pages": 28,
-      "shards": 2
+      "seconds": 2.6776525460008997,
+      "bytes": 7165592,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.7378346169971337,
-      "bytes": 10782839,
-      "pages": 36,
+      "seconds": 1.7035883720018319,
+      "bytes": 12069614,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.3063333719983348,
-      "bytes": 7915290,
+      "seconds": 1.154275818997121,
+      "bytes": 1133519,
+      "pages": 32,
+      "shards": 1
+    },
+    {
+      "seconds": 1.6667711109985248,
+      "bytes": 11320824,
+      "pages": 31,
+      "shards": 4
+    },
+    {
+      "seconds": 1.2526901849996648,
+      "bytes": 1223359,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1693511949997628,
+      "bytes": 7842479,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0522420070010412,
-      "bytes": 1136593,
-      "pages": 36,
+      "seconds": 0.9436758240008203,
+      "bytes": 855631,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.2345661709987326,
-      "bytes": 11661576,
-      "pages": 28,
-      "shards": 3
+      "seconds": 1.5082043110014638,
+      "bytes": 16602529,
+      "pages": 32,
+      "shards": 4
     },
     {
-      "seconds": 1.2733937859993603,
-      "bytes": 6517566,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.5457586960001208,
+      "bytes": 7678717,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.2599434170006134,
-      "bytes": 6272622,
+      "seconds": 0.9852680410003813,
+      "bytes": 917014,
+      "pages": 26,
+      "shards": 1
+    },
+    {
+      "seconds": 3.7047144000025582,
+      "bytes": 6068293,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.9948137850005878,
-      "bytes": 7735186,
-      "pages": 36,
+      "seconds": 1.0213589359991602,
+      "bytes": 2097340,
+      "pages": 21,
+      "shards": 2
+    },
+    {
+      "seconds": 1.7385368430004746,
+      "bytes": 10700049,
+      "pages": 47,
       "shards": 3
     },
     {
-      "seconds": 1.6032572549993347,
-      "bytes": 5165320,
-      "pages": 25,
+      "seconds": 1.4434581130008155,
+      "bytes": 11762868,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.317350670000451,
+      "bytes": 2491706,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.8941422179996152,
-      "bytes": 6019433,
-      "pages": 39,
+      "seconds": 1.2309653659976902,
+      "bytes": 5725164,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3825668060017051,
-      "bytes": 13491032,
+      "seconds": 2.086732669002231,
+      "bytes": 12324271,
       "pages": 40,
-      "shards": 5
+      "shards": 4
     },
     {
-      "seconds": 0.9105340399983106,
-      "bytes": 1249369,
-      "pages": 35,
+      "seconds": 1.4604756749977241,
+      "bytes": 1030943,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.295263830001204,
-      "bytes": 5736433,
-      "pages": 23,
+      "seconds": 1.2873325260006823,
+      "bytes": 5911958,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.218638172998908,
-      "bytes": 7112834,
-      "pages": 43,
-      "shards": 3
+      "seconds": 1.1188029909972101,
+      "bytes": 2149233,
+      "pages": 21,
+      "shards": 2
     },
     {
-      "seconds": 1.2346669860016846,
-      "bytes": 6391793,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.4335154450018308,
+      "bytes": 16606396,
+      "pages": 45,
+      "shards": 4
     },
     {
-      "seconds": 1.537398492997454,
-      "bytes": 7291338,
-      "pages": 30,
-      "shards": 3
+      "seconds": 0.9105693140008952,
+      "bytes": 1064335,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.5484525030005898,
-      "bytes": 10816659,
-      "pages": 38,
+      "seconds": 1.423822492997715,
+      "bytes": 7690011,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 2.108060000999103,
-      "bytes": 5732876,
-      "pages": 29,
+      "seconds": 1.2330543320022116,
+      "bytes": 6359348,
+      "pages": 37,
       "shards": 2
+    },
+    {
+      "seconds": 1.232538223001029,
+      "bytes": 5896962,
+      "pages": 25,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ba9bbe2eb54c264d2b9324ec47dc07ced3c51c8d254214772fea4aa3594fd98
-size 721768447
+oid sha256:b5920c2beb588bacfe2a156d706d2a8ce72852d4602bd521d016d792453dc504
+size 721778751

training_config.json

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