Qwen90M step 685000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 680000,
+  "step": 685000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "4a1d373cfe63b63f51c7756f3587c44e326964e25d196bd6fd7f16ad403624e6",
+    "model.safetensors": "2f198abaec427b39945f67e2dc4ae368c1e1f3bc749571b34df3aa6e1fabdbbc",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "def47c4f3c010dce6d7624c71b7cffcba72fe3c499691cd19b43ed26143caae4",
-    "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
+    "trainer.pt": "fef7bf8813882524f5ca30eee990ca62de8fcdd9a4aac08f179ed7a0bfb2f7cb",
+    "training_config.json": "13fe2f3f4e97181994bd5fcebc103bd2c14b5ea2ce9da3213db6d9bf51f1066f",
     "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf",
-    "shard_scheduler.json": "c51fe95c8d0f826af2d22730226b70169bd01d52e0fa35e70b3c271ed878f995",
-    "coverage.json": "1fa3ed346a97a26d94a789368927da0f3d9343e3c8dde50979494db8173cd5df",
+    "shard_scheduler.json": "fa63f6790e325f60a69dde36bb42495987d8d2fe00e2528f2132581755f7dea1",
+    "coverage.json": "327c66438413b6fcd5c9a5d61ad54b0268663e1e6f30a1e6729f0395214b9ec3",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 680000,
+  "step": 685000,
   "stats": {
-    "climbmix/blocks": 24480000,
-    "climbmix/targets": 40923747216,
-    "climbmix/input_tokens": 40926340688,
-    "ultra_qa/blocks": 9180000,
-    "ultra_qa/targets": 14976547101,
-    "ultra_qa/input_tokens": 14976576055,
-    "cortex/blocks": 9180000,
-    "cortex/targets": 349045275,
-    "cortex/input_tokens": 14592186628,
-    "ultra_style/blocks": 9180000,
-    "ultra_style/targets": 14519965635,
-    "ultra_style/input_tokens": 14520727794,
-    "openmath/blocks": 6120000,
-    "openmath/targets": 9456596693,
-    "openmath/input_tokens": 9459747141,
-    "rewrite6/blocks": 3060000,
-    "rewrite6/targets": 5016477122,
-    "rewrite6/input_tokens": 5021976869
+    "climbmix/blocks": 24660000,
+    "climbmix/targets": 41224942465,
+    "climbmix/input_tokens": 41227554983,
+    "ultra_qa/blocks": 9247500,
+    "ultra_qa/targets": 15086681728,
+    "ultra_qa/input_tokens": 15086710887,
+    "cortex/blocks": 9247500,
+    "cortex/targets": 351611655,
+    "cortex/input_tokens": 14699417911,
+    "ultra_style/blocks": 9247500,
+    "ultra_style/targets": 14626945782,
+    "ultra_style/input_tokens": 14627713723,
+    "openmath/blocks": 6165000,
+    "openmath/targets": 9526167364,
+    "openmath/input_tokens": 9529340892,
+    "rewrite6/blocks": 3082500,
+    "rewrite6/targets": 5053314593,
+    "rewrite6/input_tokens": 5058855101
   },
   "sources": {
     "climbmix": {
-      "page": 3352,
-      "row": 2082,
+      "page": 3377,
+      "row": 676,
       "start": {
-        "file": 162,
+        "file": 163,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 20,
+          "row_group": 37,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 13731874
+      "offset": 4096,
+      "documents": 13832868
     },
     "rewrite6": {
-      "page": 7,
-      "row": 1154,
+      "page": 21,
+      "row": 1829,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 28672
+          "row": 86016
         }
       },
       "epoch": 38,
       "offset": 0,
-      "documents": 7835178
+      "documents": 7893197
     },
     "openmath": {
-      "page": 2095,
-      "row": 988,
+      "page": 2110,
+      "row": 2518,
       "start": {
         "file": 19,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 284,
-          "row": 770
+          "row_group": 346,
+          "row": 210
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 8582108
+      "offset": 0,
+      "documents": 8645078
     },
     "ultra_qa": {
-      "page": 906,
-      "row": 0,
+      "page": 912,
+      "row": 2704,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 9040
+          "row": 33616
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 3710976
+      "documents": 3738256
     },
     "ultra_style": {
-      "page": 1382,
-      "row": 3920,
+      "page": 1393,
+      "row": 671,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 2,
-          "row": 11087
+          "row": 56143
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 5664592
+      "documents": 5706399
     },
     "cortex": {
-      "page": 178037,
-      "row": 110,
+      "page": 179346,
+      "row": 60,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 22788846
+      "documents": 22956348
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 680000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "4a1d373cfe63b63f51c7756f3587c44e326964e25d196bd6fd7f16ad403624e6", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "def47c4f3c010dce6d7624c71b7cffcba72fe3c499691cd19b43ed26143caae4", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf", "shard_scheduler.json": "c51fe95c8d0f826af2d22730226b70169bd01d52e0fa35e70b3c271ed878f995", "coverage.json": "1fa3ed346a97a26d94a789368927da0f3d9343e3c8dde50979494db8173cd5df", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000680000"}
+{"step": 685000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2f198abaec427b39945f67e2dc4ae368c1e1f3bc749571b34df3aa6e1fabdbbc", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "fef7bf8813882524f5ca30eee990ca62de8fcdd9a4aac08f179ed7a0bfb2f7cb", "training_config.json": "13fe2f3f4e97181994bd5fcebc103bd2c14b5ea2ce9da3213db6d9bf51f1066f", "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf", "shard_scheduler.json": "fa63f6790e325f60a69dde36bb42495987d8d2fe00e2528f2132581755f7dea1", "coverage.json": "327c66438413b6fcd5c9a5d61ad54b0268663e1e6f30a1e6729f0395214b9ec3", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000685000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a1d373cfe63b63f51c7756f3587c44e326964e25d196bd6fd7f16ad403624e6
+oid sha256:2f198abaec427b39945f67e2dc4ae368c1e1f3bc749571b34df3aa6e1fabdbbc
 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": 76877.10649082456,
-      "required_pages_per_second_upper_estimate": 0.005318536606498196
+      "required_token_positions_per_second": 75042.20393901474,
+      "required_pages_per_second_upper_estimate": 0.005191593790403495
     },
     {
       "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": 9609.63831135307,
-      "required_pages_per_second_upper_estimate": 0.004023354888896408
+      "required_token_positions_per_second": 9380.275492376843,
+      "required_pages_per_second_upper_estimate": 0.00392732546623137
     },
     {
       "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": 19219.27662270614,
-      "required_pages_per_second_upper_estimate": 0.004002868474101203
+      "required_token_positions_per_second": 18760.550984753685,
+      "required_pages_per_second_upper_estimate": 0.003907328021124296
     },
     {
       "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": 28828.91493405921,
-      "required_pages_per_second_upper_estimate": 0.0017282494310925398
+      "required_token_positions_per_second": 28140.826477130526,
+      "required_pages_per_second_upper_estimate": 0.0016869995787499043
     },
     {
       "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": 28828.91493405921,
-      "required_pages_per_second_upper_estimate": 0.0022413213311274736
+      "required_token_positions_per_second": 28140.826477130526,
+      "required_pages_per_second_upper_estimate": 0.0021878254801825318
     },
     {
       "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": 28828.91493405921,
-      "required_pages_per_second_upper_estimate": 0.36978636669692805
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28140.826477130526,
+      "required_pages_per_second_upper_estimate": 0.3609603067832702
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,412 +155,412 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.13814311317139527,
-    "producer_p95_seconds": 0.3269497630026308,
-    "gpu_update_seconds": 0.9590371355716879,
-    "producer_batches_per_second": 7.238869727507096,
-    "gpu_batches_per_second": 1.0427124903811924,
-    "producer_headroom": 6.94234488824501,
+    "producer_batch_seconds": 0.3391172111871015,
+    "producer_p95_seconds": 0.355980565000209,
+    "gpu_update_seconds": 0.9824871356379303,
+    "producer_batches_per_second": 2.948832931538438,
+    "gpu_batches_per_second": 1.0178250317249178,
+    "producer_headroom": 2.897190420382001,
     "sustainable": true,
-    "buffer_seconds": 60.419339541016335,
+    "buffer_seconds": 60.91420240955168,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 548419.5468379183,
-    "upload_bytes_per_second": 5461558.934611099,
+    "generation_bytes_per_second": 385732.1530428105,
+    "upload_bytes_per_second": 4949797.310566265,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 117.67806819775737,
+    "estimated_full_shards_per_hour": 82.769140658028,
     "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": 1.07342461799999,
-      "bytes": 5972281,
-      "pages": 29,
+      "seconds": 1.6456713849984226,
+      "bytes": 6071875,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.2648427659987647,
-      "bytes": 7742666,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1688129319991276,
-      "bytes": 5401350,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.52896361799867,
+      "bytes": 13388189,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.2462927969972952,
-      "bytes": 6148998,
-      "pages": 33,
+      "seconds": 1.3202310080014286,
+      "bytes": 5863529,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.1124823669997568,
-      "bytes": 8182028,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1143527310014179,
-      "bytes": 6164100,
+      "seconds": 1.5651336049995734,
+      "bytes": 6044120,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3281419349987118,
-      "bytes": 11547797,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8457415819975722,
-      "bytes": 905737,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.1207633359990723,
+      "bytes": 2404167,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.6407027059976826,
-      "bytes": 6090920,
-      "pages": 36,
+      "seconds": 1.4597645020003256,
+      "bytes": 6256363,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0669106919995102,
-      "bytes": 2526136,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.2800013509986456,
+      "bytes": 667279,
+      "pages": 18,
+      "shards": 1
     },
     {
-      "seconds": 1.477937416999339,
-      "bytes": 16111480,
-      "pages": 34,
+      "seconds": 1.4021912890020758,
+      "bytes": 11567150,
+      "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 1.2388989300015965,
-      "bytes": 7565588,
-      "pages": 40,
+      "seconds": 1.280043183000089,
+      "bytes": 12206708,
+      "pages": 48,
       "shards": 3
     },
     {
-      "seconds": 0.9064518010018219,
-      "bytes": 1123131,
-      "pages": 35,
-      "shards": 1
+      "seconds": 0.8664777250014595,
+      "bytes": 2384518,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.0333158499997808,
-      "bytes": 5705052,
+      "seconds": 1.027337190000253,
+      "bytes": 5987324,
       "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 1.4537645320015145,
-      "bytes": 12446484,
-      "pages": 45,
-      "shards": 4
+      "seconds": 1.2340771259987378,
+      "bytes": 10655993,
+      "pages": 48,
+      "shards": 3
     },
     {
-      "seconds": 1.027312460999383,
-      "bytes": 1095320,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.3521600980020594,
+      "bytes": 2290704,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.1021007999988797,
-      "bytes": 11824886,
-      "pages": 33,
+      "seconds": 1.766581726995355,
+      "bytes": 12101436,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.0116616440027428,
-      "bytes": 2398524,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.0664826619977248,
+      "bytes": 6396203,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.1044605290007894,
-      "bytes": 9770514,
+      "seconds": 1.070789342003991,
+      "bytes": 6236244,
       "pages": 37,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.1610946500004502,
-      "bytes": 7427357,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.5840121689980151,
+      "bytes": 6547762,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3267378080017806,
-      "bytes": 1199966,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.2987593140060198,
+      "bytes": 2393188,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.69540837599925,
-      "bytes": 9973657,
-      "pages": 31,
+      "seconds": 1.808516741002677,
+      "bytes": 10401411,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.1572326720015553,
-      "bytes": 13287961,
-      "pages": 34,
+      "seconds": 1.1967187990012462,
+      "bytes": 13416312,
+      "pages": 42,
       "shards": 4
     },
     {
-      "seconds": 0.7943182929993782,
-      "bytes": 1190422,
-      "pages": 36,
+      "seconds": 0.995916217994818,
+      "bytes": 1003302,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.1334048999997322,
-      "bytes": 5665943,
-      "pages": 19,
+      "seconds": 0.9674175760010257,
+      "bytes": 6098431,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.1813513000015519,
-      "bytes": 7573359,
-      "pages": 41,
+      "seconds": 1.0489661490064464,
+      "bytes": 6157279,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.4761009860012564,
-      "bytes": 5975966,
-      "pages": 28,
+      "seconds": 1.4413105299972813,
+      "bytes": 6425996,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.119130936000147,
-      "bytes": 7848522,
-      "pages": 40,
+      "seconds": 1.128546699997969,
+      "bytes": 5952550,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.0408579100003408,
-      "bytes": 6882526,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.3368553659965983,
+      "bytes": 12766871,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.0372045380026975,
-      "bytes": 6042585,
-      "pages": 32,
+      "seconds": 1.6370693080025376,
+      "bytes": 5715454,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.5961815709997609,
-      "bytes": 6520796,
+      "seconds": 1.2990395020024152,
+      "bytes": 2748829,
       "pages": 38,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 0.9840364880001289,
-      "bytes": 6376962,
+      "seconds": 1.300822404999053,
+      "bytes": 6046016,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2160047779980232,
-      "bytes": 7548487,
-      "pages": 31,
-      "shards": 4
+      "seconds": 1.4766436999998405,
+      "bytes": 10238885,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.0057136780014844,
-      "bytes": 5529362,
-      "pages": 38,
+      "seconds": 1.4429230239984463,
+      "bytes": 2385665,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.374610462000419,
-      "bytes": 11892118,
-      "pages": 33,
+      "seconds": 1.539863748002972,
+      "bytes": 11993569,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 1.2482218100012688,
-      "bytes": 2427794,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.4481774710002355,
+      "bytes": 7324064,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.1882967479978106,
-      "bytes": 6156483,
-      "pages": 38,
+      "seconds": 1.494102450000355,
+      "bytes": 5356034,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.4572524590003013,
-      "bytes": 10153840,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.1517664069979219,
+      "bytes": 6283260,
+      "pages": 38,
+      "shards": 2
     },
     {
-      "seconds": 0.8581832199997734,
-      "bytes": 2650389,
-      "pages": 37,
+      "seconds": 1.5029984030043124,
+      "bytes": 2302830,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.6456713849984226,
-      "bytes": 6071875,
+      "seconds": 1.5264721800049301,
+      "bytes": 6345437,
       "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.52896361799867,
-      "bytes": 13388189,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.7716346369998064,
+      "bytes": 17764299,
+      "pages": 34,
+      "shards": 5
     },
     {
-      "seconds": 1.3202310080014286,
-      "bytes": 5863529,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.4270490490016527,
+      "bytes": 1152119,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.5651336049995734,
-      "bytes": 6044120,
+      "seconds": 1.5365510039991932,
+      "bytes": 6323603,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1207633359990723,
-      "bytes": 2404167,
-      "pages": 29,
+      "seconds": 1.5189877529992373,
+      "bytes": 7424560,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2933890700005577,
+      "bytes": 4945280,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.4597645020003256,
-      "bytes": 6256363,
+      "seconds": 1.5587586719993851,
+      "bytes": 6311746,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2800013509986456,
-      "bytes": 667279,
-      "pages": 18,
-      "shards": 1
+      "seconds": 1.5925084610062186,
+      "bytes": 13513629,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 1.4021912890020758,
-      "bytes": 11567150,
-      "pages": 32,
-      "shards": 4
+      "seconds": 1.9457880960035254,
+      "bytes": 1015286,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.280043183000089,
-      "bytes": 12206708,
-      "pages": 48,
-      "shards": 3
+      "seconds": 1.624823366997589,
+      "bytes": 11605166,
+      "pages": 31,
+      "shards": 4
     },
     {
-      "seconds": 0.8664777250014595,
-      "bytes": 2384518,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.4567563370001153,
+      "bytes": 1138092,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.027337190000253,
-      "bytes": 5987324,
-      "pages": 24,
+      "seconds": 1.3836040919995867,
+      "bytes": 6235413,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2340771259987378,
-      "bytes": 10655993,
-      "pages": 48,
+      "seconds": 1.6395168350063614,
+      "bytes": 7620206,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.3521600980020594,
-      "bytes": 2290704,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.766581726995355,
-      "bytes": 12101436,
-      "pages": 41,
+      "seconds": 1.4781058270018548,
+      "bytes": 11305199,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0664826619977248,
-      "bytes": 6396203,
+      "seconds": 1.3714339009966352,
+      "bytes": 7352671,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.070789342003991,
-      "bytes": 6236244,
-      "pages": 37,
+      "seconds": 1.3506424060033169,
+      "bytes": 6203770,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.5840121689980151,
-      "bytes": 6547762,
+      "seconds": 1.3150392690004082,
+      "bytes": 914685,
+      "pages": 26,
+      "shards": 1
+    },
+    {
+      "seconds": 1.6476991579984315,
+      "bytes": 12016198,
+      "pages": 39,
+      "shards": 4
+    },
+    {
+      "seconds": 1.4772214850017917,
+      "bytes": 1146549,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2987593140060198,
-      "bytes": 2393188,
+      "seconds": 1.649438596003165,
+      "bytes": 11880426,
       "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9699813629995333,
+      "bytes": 2462244,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.808516741002677,
-      "bytes": 10401411,
-      "pages": 34,
+      "seconds": 1.3995045109986677,
+      "bytes": 10981344,
+      "pages": 45,
       "shards": 3
     },
     {
-      "seconds": 1.1967187990012462,
-      "bytes": 13416312,
-      "pages": 42,
-      "shards": 4
+      "seconds": 1.094142590998672,
+      "bytes": 7571723,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 0.995916217994818,
-      "bytes": 1003302,
-      "pages": 27,
+      "seconds": 1.1913003890003893,
+      "bytes": 1341787,
+      "pages": 37,
       "shards": 1
     },
     {
-      "seconds": 0.9674175760010257,
-      "bytes": 6098431,
-      "pages": 38,
+      "seconds": 1.4330705320026027,
+      "bytes": 5823183,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.0489661490064464,
-      "bytes": 6157279,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.3540045269983239,
+      "bytes": 17287207,
+      "pages": 44,
+      "shards": 5
     },
     {
-      "seconds": 1.4413105299972813,
-      "bytes": 6425996,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.2279312789978576,
+      "bytes": 1252954,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.128546699997969,
-      "bytes": 5952550,
-      "pages": 28,
+      "seconds": 1.2454847960034385,
+      "bytes": 7379042,
+      "pages": 32,
       "shards": 3
     }
   ],

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:def47c4f3c010dce6d7624c71b7cffcba72fe3c499691cd19b43ed26143caae4
-size 721786943
+oid sha256:fef7bf8813882524f5ca30eee990ca62de8fcdd9a4aac08f179ed7a0bfb2f7cb
+size 721756031

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,