Qwen90M step 870000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 865000,
+  "step": 870000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387",
+    "model.safetensors": "05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9",
-    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
+    "trainer.pt": "ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed",
+    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367",
-    "shard_scheduler.json": "c95b4d546c5df7ee17ed03b0bd9cd6934d744393d9e9e9817ef0f64a430ddf94",
-    "coverage.json": "3109f65091eb186c8eaa80402ee04bc03abb10fd556167b2666aaef70ee57e60",
+    "shard_scheduler.json": "328e638e7ad91d39ab501d496c26d65bbb1a3067cdf2542ba84f30916f4660be",
+    "coverage.json": "7d047105e0eebd85909b8ec68f55ac31224c6dec9049175243e510c634cd4bd3",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 865000,
+  "step": 870000,
   "stats": {
-    "climbmix/blocks": 31140000,
-    "climbmix/targets": 52055877655,
-    "climbmix/input_tokens": 52059176107,
-    "ultra_qa/blocks": 11677500,
-    "ultra_qa/targets": 19051109408,
-    "ultra_qa/input_tokens": 19051146044,
-    "cortex/blocks": 11677500,
-    "cortex/targets": 444043754,
-    "cortex/input_tokens": 18562155884,
-    "ultra_style/blocks": 11677500,
-    "ultra_style/targets": 18470831928,
-    "ultra_style/input_tokens": 18471800000,
-    "openmath/blocks": 7785000,
-    "openmath/targets": 12029146241,
-    "openmath/input_tokens": 12033154511,
-    "rewrite6/blocks": 3892500,
-    "rewrite6/targets": 6381217228,
-    "rewrite6/input_tokens": 6388213504
+    "climbmix/blocks": 31320000,
+    "climbmix/targets": 52356514020,
+    "climbmix/input_tokens": 52359831944,
+    "ultra_qa/blocks": 11745000,
+    "ultra_qa/targets": 19161264699,
+    "ultra_qa/input_tokens": 19161301550,
+    "cortex/blocks": 11745000,
+    "cortex/targets": 446591895,
+    "cortex/input_tokens": 18669449496,
+    "ultra_style/blocks": 11745000,
+    "ultra_style/targets": 18577453797,
+    "ultra_style/input_tokens": 18578427724,
+    "openmath/blocks": 7830000,
+    "openmath/targets": 12098676149,
+    "openmath/input_tokens": 12102707529,
+    "rewrite6/blocks": 3915000,
+    "rewrite6/targets": 6418117328,
+    "rewrite6/input_tokens": 6425153910
   },
   "sources": {
     "climbmix": {
-      "page": 4264,
-      "row": 2484,
+      "page": 4289,
+      "row": 1766,
       "start": {
-        "file": 206,
+        "file": 207,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 33,
+          "row_group": 51,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 17467828
+      "offset": 0,
+      "documents": 17569510
     },
     "rewrite6": {
-      "page": 26,
-      "row": 1330,
+      "page": 40,
+      "row": 1476,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 106496
+          "row": 163840
         }
       },
       "epoch": 48,
-      "offset": 2048,
-      "documents": 9967218
+      "offset": 0,
+      "documents": 10024708
     },
     "openmath": {
-      "page": 2665,
-      "row": 1871,
+      "page": 2680,
+      "row": 3506,
       "start": {
-        "file": 24,
+        "file": 25,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 436,
-          "row": 241
+          "row_group": 61,
+          "row": 32
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 10917711
+      "offset": 0,
+      "documents": 10980786
     },
     "ultra_qa": {
-      "page": 1152,
-      "row": 1712,
+      "page": 1159,
+      "row": 365,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 0,
-          "row": 41950
+          "row_group": 1,
+          "row": 6079
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 4720304
+      "offset": 4096,
+      "documents": 4747629
     },
     "ultra_style": {
-      "page": 1758,
-      "row": 2694,
+      "page": 1768,
+      "row": 3754,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 12042
+          "row": 53002
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 7203462
+      "offset": 2048,
+      "documents": 7245482
     },
     "cortex": {
-      "page": 226468,
-      "row": 78,
+      "page": 227777,
+      "row": 116,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 28987982
+      "documents": 29155572
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 865000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9", "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "c95b4d546c5df7ee17ed03b0bd9cd6934d744393d9e9e9817ef0f64a430ddf94", "coverage.json": "3109f65091eb186c8eaa80402ee04bc03abb10fd556167b2666aaef70ee57e60", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000865000"}
+{"step": 870000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "328e638e7ad91d39ab501d496c26d65bbb1a3067cdf2542ba84f30916f4660be", "coverage.json": "7d047105e0eebd85909b8ec68f55ac31224c6dec9049175243e510c634cd4bd3", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000870000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387
+oid sha256:05b839c3aab35c30b88ee03ce1a7eb869f4818cc4eb81eb58c68d7bf3452de5b
 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": 75711.298881415,
-      "required_pages_per_second_upper_estimate": 0.0052378833310329055
+      "required_token_positions_per_second": 77384.34908444676,
+      "required_pages_per_second_upper_estimate": 0.005353628826089951
     },
     {
       "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": 9463.912360176875,
-      "required_pages_per_second_upper_estimate": 0.003962342476242838
+      "required_token_positions_per_second": 9673.043635555845,
+      "required_pages_per_second_upper_estimate": 0.004049901374086377
     },
     {
       "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": 18927.82472035375,
-      "required_pages_per_second_upper_estimate": 0.003942166728944734
+      "required_token_positions_per_second": 19346.08727111169,
+      "required_pages_per_second_upper_estimate": 0.004029279787942392
     },
     {
       "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": 28391.737080530624,
-      "required_pages_per_second_upper_estimate": 0.0017020412863054321
+      "required_token_positions_per_second": 29019.130906667535,
+      "required_pages_per_second_upper_estimate": 0.001739652588207436
     },
     {
       "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": 28391.737080530624,
-      "required_pages_per_second_upper_estimate": 0.002207332675957781
+      "required_token_positions_per_second": 29019.130906667535,
+      "required_pages_per_second_upper_estimate": 0.0022561097863261297
     },
     {
       "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": 28391.737080530624,
-      "required_pages_per_second_upper_estimate": 0.3641787185968705
+      "estimated_pages_for_buffer": 23,
+      "required_token_positions_per_second": 29019.130906667535,
+      "required_pages_per_second_upper_estimate": 0.37222625295554873
     }
   ],
   "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": 0.13520461964066044,
-    "producer_p95_seconds": 0.2855433209988405,
-    "gpu_update_seconds": 0.9738044530906622,
-    "producer_batches_per_second": 7.396196983932548,
-    "gpu_batches_per_second": 1.0269002126928033,
-    "producer_headroom": 7.202449558889239,
+    "producer_batch_seconds": 0.10117103117187298,
+    "producer_p95_seconds": 0.2711704980029026,
+    "gpu_update_seconds": 0.9527507935686489,
+    "producer_batches_per_second": 9.884252324177304,
+    "gpu_batches_per_second": 1.049592408371945,
+    "producer_headroom": 9.417229245692688,
     "sustainable": true,
-    "buffer_seconds": 60.37587609162106,
+    "buffer_seconds": 60.02329999482488,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 77,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 358001.55127571075,
-    "upload_bytes_per_second": 3629836.3380909152,
+    "generation_bytes_per_second": 322491.0104688514,
+    "upload_bytes_per_second": 4817943.822503319,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 76.81879905417911,
+    "estimated_full_shards_per_hour": 69.1990636401096,
     "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": 1.716705255999841,
-      "bytes": 6235687,
+      "seconds": 0.990031227000145,
+      "bytes": 1215408,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.4002689310000278,
+      "bytes": 11880783,
+      "pages": 32,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1156020849994093,
+      "bytes": 6256874,
+      "pages": 38,
+      "shards": 2
+    },
+    {
+      "seconds": 1.0478186670006835,
+      "bytes": 6856744,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 33.88602663399979,
-      "bytes": 7355362,
-      "pages": 31,
+      "seconds": 1.187728501001402,
+      "bytes": 7914573,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.5573600560001069,
-      "bytes": 11648033,
-      "pages": 46,
-      "shards": 3
+      "seconds": 1.098511371998029,
+      "bytes": 5354193,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.2928704839996499,
-      "bytes": 7315946,
+      "seconds": 2.7016539569995075,
+      "bytes": 7555720,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.4081739580014982,
-      "bytes": 5891309,
+      "seconds": 1.483174205000978,
+      "bytes": 5900678,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.286470338000072,
-      "bytes": 2778002,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.8595360259969311,
+      "bytes": 1182793,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.1679330120005034,
-      "bytes": 10820666,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.7856543759990018,
+      "bytes": 17299146,
+      "pages": 27,
+      "shards": 6
     },
     {
-      "seconds": 1.4827527400011604,
-      "bytes": 6647922,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.9336543670033279,
+      "bytes": 1145946,
+      "pages": 36,
+      "shards": 1
     },
     {
-      "seconds": 1.248109399999521,
-      "bytes": 7422482,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.1468644119995588,
+      "bytes": 6284755,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.451942487999986,
-      "bytes": 6077412,
-      "pages": 30,
+      "seconds": 1.0925358679996862,
+      "bytes": 2671177,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4631003799986502,
-      "bytes": 7075109,
-      "pages": 43,
+      "seconds": 1.42488349499763,
+      "bytes": 9837963,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 2.129564337999909,
-      "bytes": 5884538,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.2818151600004057,
+      "bytes": 7876217,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.1728028199995606,
-      "bytes": 6147166,
-      "pages": 33,
+      "seconds": 1.1551925370004028,
+      "bytes": 6771836,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2317090710002958,
-      "bytes": 8202394,
-      "pages": 32,
+      "seconds": 1.460754736999661,
+      "bytes": 10771768,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.6838970150001842,
-      "bytes": 9992623,
-      "pages": 38,
-      "shards": 3
+      "seconds": 0.9620606520002184,
+      "bytes": 2617341,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.2029927740004496,
-      "bytes": 6068424,
+      "seconds": 1.1643879789990024,
+      "bytes": 6176643,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1870869580016006,
-      "bytes": 2666964,
-      "pages": 36,
+      "seconds": 1.7241964509994432,
+      "bytes": 5999571,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.4462591800001974,
-      "bytes": 10549433,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0944702749984572,
-      "bytes": 2560030,
-      "pages": 36,
+      "seconds": 1.129534389001492,
+      "bytes": 2663790,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3531785280028998,
-      "bytes": 11787584,
-      "pages": 32,
+      "seconds": 1.8240933249981026,
+      "bytes": 16162613,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 1.0967427190007584,
-      "bytes": 6096101,
-      "pages": 36,
+      "seconds": 1.3452121949994762,
+      "bytes": 2460437,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3933399300003657,
-      "bytes": 2445888,
-      "pages": 28,
+      "seconds": 1.3531316250009695,
+      "bytes": 6148520,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.9226194889997714,
-      "bytes": 9744014,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.9429252529989753,
+      "bytes": 341818,
+      "pages": 9,
+      "shards": 1
     },
     {
-      "seconds": 1.1809931960015092,
-      "bytes": 7571529,
-      "pages": 39,
+      "seconds": 1.6665152259993192,
+      "bytes": 10630836,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.7855844450023142,
-      "bytes": 302035,
-      "pages": 9,
-      "shards": 1
+      "seconds": 1.7391449930000817,
+      "bytes": 7296963,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 0.93341918599981,
-      "bytes": 1061744,
-      "pages": 32,
+      "seconds": 1.3247330819976924,
+      "bytes": 862254,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.5622991370000818,
-      "bytes": 16646339,
-      "pages": 40,
+      "seconds": 1.7279805650032358,
+      "bytes": 13248087,
+      "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 1.5330627459989046,
-      "bytes": 7305115,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8943780709996645,
-      "bytes": 971661,
-      "pages": 29,
-      "shards": 1
+      "seconds": 1.2414953469997272,
+      "bytes": 4717232,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.4041120760011836,
-      "bytes": 7469317,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.1668012100017222,
+      "bytes": 6253390,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.2255340310002794,
-      "bytes": 10181330,
-      "pages": 31,
+      "seconds": 1.4145583839999745,
+      "bytes": 7569385,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0713620990027266,
-      "bytes": 1066838,
+      "seconds": 1.1982823690013902,
+      "bytes": 1172595,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.2431204950007668,
-      "bytes": 13191953,
-      "pages": 36,
+      "seconds": 1.8298632039986842,
+      "bytes": 11926092,
+      "pages": 29,
       "shards": 4
     },
     {
-      "seconds": 1.2758062670000072,
-      "bytes": 1083353,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.4296780900003796,
+      "bytes": 6878310,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.7504430440021679,
-      "bytes": 10417035,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.173846097997739,
+      "bytes": 6578162,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.7296301900023536,
-      "bytes": 7708812,
-      "pages": 39,
+      "seconds": 1.273192676002509,
+      "bytes": 7543138,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.0786898769983964,
-      "bytes": 798109,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.0804058459980297,
+      "bytes": 5227494,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.6809108209999977,
-      "bytes": 7259646,
-      "pages": 29,
+      "seconds": 1.174932928999624,
+      "bytes": 7282353,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.6123849089999567,
-      "bytes": 16470150,
-      "pages": 45,
-      "shards": 4
-    },
-    {
-      "seconds": 1.350964933000796,
-      "bytes": 1228633,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3665898539984482,
+      "bytes": 6068190,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.590518836001138,
-      "bytes": 7289840,
-      "pages": 30,
+      "seconds": 1.6203178860014305,
+      "bytes": 6721657,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 0.990031227000145,
-      "bytes": 1215408,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4002689310000278,
-      "bytes": 11880783,
+      "seconds": 1.8514023890020326,
+      "bytes": 12113021,
       "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 1.1156020849994093,
-      "bytes": 6256874,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.0979482909970102,
+      "bytes": 1151793,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.0478186670006835,
-      "bytes": 6856744,
-      "pages": 36,
+      "seconds": 1.1090724049972778,
+      "bytes": 6289086,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.187728501001402,
-      "bytes": 7914573,
-      "pages": 30,
+      "seconds": 1.7820321440012776,
+      "bytes": 7365760,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.098511371998029,
-      "bytes": 5354193,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.5119859270016605,
+      "bytes": 1025135,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 2.7016539569995075,
-      "bytes": 7555720,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.464499204997992,
+      "bytes": 17226254,
+      "pages": 41,
+      "shards": 5
     },
     {
-      "seconds": 1.483174205000978,
-      "bytes": 5900678,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.9209507179984939,
+      "bytes": 949705,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 0.8595360259969311,
-      "bytes": 1182793,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.9317821920012648,
+      "bytes": 6082541,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.7856543759990018,
-      "bytes": 17299146,
-      "pages": 27,
-      "shards": 6
+      "seconds": 1.9298742110004241,
+      "bytes": 11871053,
+      "pages": 33,
+      "shards": 4
     },
     {
-      "seconds": 0.9336543670033279,
-      "bytes": 1145946,
+      "seconds": 1.752713240999583,
+      "bytes": 1247797,
       "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.1468644119995588,
-      "bytes": 6284755,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.4635059559986985,
+      "bytes": 7719999,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.0925358679996862,
-      "bytes": 2671177,
-      "pages": 37,
+      "seconds": 1.2099752760004776,
+      "bytes": 6423731,
+      "pages": 22,
       "shards": 2
     },
     {
-      "seconds": 1.42488349499763,
-      "bytes": 9837963,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2818151600004057,
-      "bytes": 7876217,
+      "seconds": 1.4774333919995115,
+      "bytes": 10519718,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1551925370004028,
-      "bytes": 6771836,
+      "seconds": 1.5275958259990148,
+      "bytes": 7252352,
       "pages": 30,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.460754736999661,
-      "bytes": 10771768,
+      "seconds": 1.0232120440014114,
+      "bytes": 1138613,
       "pages": 35,
-      "shards": 3
+      "shards": 1
     },
     {
-      "seconds": 0.9620606520002184,
-      "bytes": 2617341,
-      "pages": 35,
+      "seconds": 1.0960563930020726,
+      "bytes": 6111629,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.1643879789990024,
-      "bytes": 6176643,
-      "pages": 37,
+      "seconds": 1.3281459399986488,
+      "bytes": 11209095,
+      "pages": 33,
+      "shards": 4
+    },
+    {
+      "seconds": 1.3028533390024677,
+      "bytes": 6867913,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.7241964509994432,
-      "bytes": 5999571,
-      "pages": 29,
+      "seconds": 1.275760786000319,
+      "bytes": 7467840,
+      "pages": 35,
+      "shards": 3
+    },
+    {
+      "seconds": 1.235418092001055,
+      "bytes": 6152144,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.129534389001492,
-      "bytes": 2663790,
-      "pages": 37,
+      "seconds": 1.3146963539984426,
+      "bytes": 5075870,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.8240933249981026,
-      "bytes": 16162613,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.278188054999191,
+      "bytes": 7607915,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.3452121949994762,
-      "bytes": 2460437,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.058052360000147,
+      "bytes": 885421,
+      "pages": 27,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9
-size 721784575
+oid sha256:ccf573d492eb2406d088a78d4e25d0fb9246c301889627aa9ad1a9d93fbfb1ed
+size 721769279

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,