Qwen90M step 910000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 905000,
+  "step": 910000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630",
+    "model.safetensors": "b4df0ef53a5a343fee1e4dce6ab8a731c5f01c993a900224ba917205158c612a",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4",
-    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
+    "trainer.pt": "8ba9bbe2eb54c264d2b9324ec47dc07ced3c51c8d254214772fea4aa3594fd98",
+    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
     "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "b24e87ee5310f8cae440ce163c660971b4bd3a3c4202793a1fd52e7a8b911217",
-    "coverage.json": "2b0b0982d723d593e8b8cd110f0581bc41bb25cbc26ab9aa84796b740f5c7b28",
+    "shard_scheduler.json": "dd11a2928b6fc6dbcfc1eafd5d86aa23cf716d5b4273f83412bf2508c81e3964",
+    "coverage.json": "26aa8059e8f72378a7c5a4bd2300bf3c7cff681ad595ec96d08d6b092bc45174",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 905000,
+  "step": 910000,
   "stats": {
-    "climbmix/blocks": 32580000,
-    "climbmix/targets": 54463088890,
-    "climbmix/input_tokens": 54466540635,
-    "ultra_qa/blocks": 12217500,
-    "ultra_qa/targets": 19931980030,
-    "ultra_qa/input_tokens": 19932018303,
-    "cortex/blocks": 12217500,
-    "cortex/targets": 464554652,
-    "cortex/input_tokens": 19420649920,
-    "ultra_style/blocks": 12217500,
-    "ultra_style/targets": 19324167416,
-    "ultra_style/input_tokens": 19325180301,
-    "openmath/blocks": 8145000,
-    "openmath/targets": 12585378739,
-    "openmath/input_tokens": 12589571832,
-    "rewrite6/blocks": 4072500,
-    "rewrite6/targets": 6676302372,
-    "rewrite6/input_tokens": 6683622035
+    "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
   },
   "sources": {
     "climbmix": {
-      "page": 4462,
-      "row": 661,
+      "page": 4486,
+      "row": 2973,
       "start": {
-        "file": 215,
+        "file": 217,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 82,
+          "row_group": 13,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 18277013
+      "documents": 18377629
     },
     "rewrite6": {
-      "page": 38,
-      "row": 2160,
+      "page": 2,
+      "row": 1549,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 155648
+          "row": 8192
         }
       },
-      "epoch": 50,
-      "offset": 0,
-      "documents": 10428008
+      "epoch": 51,
+      "offset": 2048,
+      "documents": 10485345
     },
     "openmath": {
-      "page": 2788,
-      "row": 2605,
+      "page": 2804,
+      "row": 458,
       "start": {
         "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 66,
-          "row": 751
+          "row_group": 132,
+          "row": 287
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11422253
+      "documents": 11485642
     },
     "ultra_qa": {
-      "page": 1205,
-      "row": 2927,
+      "page": 1212,
+      "row": 1525,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 744
+          "row": 29416
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4938607
+      "documents": 4965877
     },
     "ultra_style": {
-      "page": 1840,
-      "row": 498,
+      "page": 1850,
+      "row": 2060,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 2284
+          "row": 43244
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 7537138
+      "documents": 7579660
     },
     "cortex": {
-      "page": 236942,
-      "row": 28,
+      "page": 238252,
+      "row": 11,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 30328604
+      "documents": 30496267
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 905000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "b24e87ee5310f8cae440ce163c660971b4bd3a3c4202793a1fd52e7a8b911217", "coverage.json": "2b0b0982d723d593e8b8cd110f0581bc41bb25cbc26ab9aa84796b740f5c7b28", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000905000"}
+{"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"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630
+oid sha256:b4df0ef53a5a343fee1e4dce6ab8a731c5f01c993a900224ba917205158c612a
 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": 76565.79414645793,
-      "required_pages_per_second_upper_estimate": 0.005296999296170754
+      "required_token_positions_per_second": 76105.76941737354,
+      "required_pages_per_second_upper_estimate": 0.0052651737180082646
     },
     {
       "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": 9570.724268307242,
-      "required_pages_per_second_upper_estimate": 0.0040070623916907445
+      "required_token_positions_per_second": 9513.221177171692,
+      "required_pages_per_second_upper_estimate": 0.00398298704823338
     },
     {
       "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": 19141.448536614484,
-      "required_pages_per_second_upper_estimate": 0.003986658936232964
+      "required_token_positions_per_second": 19026.442354343384,
+      "required_pages_per_second_upper_estimate": 0.003962706181382876
     },
     {
       "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": 28712.172804921724,
-      "required_pages_per_second_upper_estimate": 0.0017212509187056575
+      "required_token_positions_per_second": 28539.663531515074,
+      "required_pages_per_second_upper_estimate": 0.0017109092511713377
     },
     {
       "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": 28712.172804921724,
-      "required_pages_per_second_upper_estimate": 0.0022322451440813925
+      "required_token_positions_per_second": 28539.663531515074,
+      "required_pages_per_second_upper_estimate": 0.002218833306862128
     },
     {
       "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": 28712.172804921724,
-      "required_pages_per_second_upper_estimate": 0.36828892401228464
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28539.663531515074,
+      "required_pages_per_second_upper_estimate": 0.3660761602790507
     }
   ],
   "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": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.067691734976421,
-    "producer_p95_seconds": 0.24887008000223432,
-    "gpu_update_seconds": 0.9629365282748887,
-    "producer_batches_per_second": 14.772852259560626,
-    "gpu_batches_per_second": 1.03849004647431,
-    "producer_headroom": 14.225319067539154,
+    "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,
     "sustainable": true,
-    "buffer_seconds": 60.66500128131799,
+    "buffer_seconds": 60.062936555194916,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 346296.08435846475,
-    "upload_bytes_per_second": 5025204.241395261,
+    "generation_bytes_per_second": 371269.35917070275,
+    "upload_bytes_per_second": 4902520.230367588,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 74.30707834306199,
+    "estimated_full_shards_per_hour": 79.66576176968395,
     "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": 1.3964128229999915,
-      "bytes": 12688722,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.3777095789992018,
+      "bytes": 11141641,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.4358708510008,
-      "bytes": 6164418,
-      "pages": 37,
+      "seconds": 3.434384232998127,
+      "bytes": 5971898,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5897931969993806,
-      "bytes": 7707081,
-      "pages": 36,
+      "seconds": 1.1304158860002644,
+      "bytes": 8230136,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.0018968719996337,
-      "bytes": 927002,
-      "pages": 29,
-      "shards": 1
+      "seconds": 1.2011196589992323,
+      "bytes": 6075797,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.3852189129993349,
-      "bytes": 10288556,
-      "pages": 35,
+      "seconds": 1.276424440999108,
+      "bytes": 6070440,
+      "pages": 25,
       "shards": 3
     },
     {
-      "seconds": 1.2743636189989047,
-      "bytes": 2631065,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.2386367050021363,
+      "bytes": 6009209,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.369053440999778,
-      "bytes": 11976832,
+      "seconds": 1.1234037470021576,
+      "bytes": 6172938,
       "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3680221860013262,
-      "bytes": 5930978,
-      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.3103218420001213,
-      "bytes": 6746870,
+      "seconds": 0.8850959770024929,
+      "bytes": 2438669,
       "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4409255560003658,
-      "bytes": 6062763,
-      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1668520560015168,
-      "bytes": 7696896,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1897505889992317,
-      "bytes": 1090330,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.6140179690009973,
+      "bytes": 15868305,
+      "pages": 31,
+      "shards": 4
     },
     {
-      "seconds": 1.3477415189991007,
-      "bytes": 15713674,
+      "seconds": 1.348259312002483,
+      "bytes": 5928296,
       "pages": 35,
-      "shards": 4
+      "shards": 2
     },
     {
-      "seconds": 1.2074203040010616,
-      "bytes": 7463350,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.9825653119987692,
+      "bytes": 2330980,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.190478869000799,
-      "bytes": 1226986,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2888210940000135,
+      "bytes": 6181272,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.5480925499996374,
-      "bytes": 11721285,
-      "pages": 32,
-      "shards": 4
+      "seconds": 1.1400288160002674,
+      "bytes": 6499509,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.0147747470000468,
-      "bytes": 1162064,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1078390350012342,
+      "bytes": 6180333,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.132629542000359,
-      "bytes": 6067377,
-      "pages": 33,
+      "seconds": 1.1313278150009864,
+      "bytes": 6410101,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.3739411900005507,
-      "bytes": 13347857,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.2548322069997084,
+      "bytes": 7827361,
+      "pages": 41,
+      "shards": 3
     },
     {
-      "seconds": 1.5033581300012884,
-      "bytes": 979335,
-      "pages": 32,
-      "shards": 1
+      "seconds": 1.3341312570009904,
+      "bytes": 10704029,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.363758608000353,
-      "bytes": 11974676,
-      "pages": 44,
-      "shards": 4
+      "seconds": 1.136130429997138,
+      "bytes": 965107,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.1600040840003203,
-      "bytes": 6119662,
+      "seconds": 1.3451466709993838,
+      "bytes": 7707115,
       "pages": 36,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 0.8502458040002239,
-      "bytes": 891680,
-      "pages": 27,
-      "shards": 1
+      "seconds": 2.3437790619973384,
+      "bytes": 11207805,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.5061208509996504,
-      "bytes": 11821724,
-      "pages": 33,
-      "shards": 4
+      "seconds": 2.2792270009995264,
+      "bytes": 8233418,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.1246318519988563,
-      "bytes": 6660673,
-      "pages": 27,
+      "seconds": 1.1861695589977899,
+      "bytes": 6023568,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2537715590005973,
-      "bytes": 5914249,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.0451825370000734,
+      "bytes": 965218,
+      "pages": 28,
+      "shards": 1
     },
     {
-      "seconds": 1.3754841779991693,
-      "bytes": 2754086,
+      "seconds": 1.2396269109995046,
+      "bytes": 7721145,
       "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.24713432200042,
-      "bytes": 5975341,
-      "pages": 36,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.473726465999789,
-      "bytes": 9735362,
-      "pages": 31,
+      "seconds": 1.3206068420004158,
+      "bytes": 11244076,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1486626710011478,
-      "bytes": 2585051,
-      "pages": 37,
+      "seconds": 1.1782944210026471,
+      "bytes": 2347277,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.5323144779995346,
-      "bytes": 11971610,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.2341140790013014,
+      "bytes": 11781395,
+      "pages": 31,
+      "shards": 4
     },
     {
-      "seconds": 1.079354624998814,
-      "bytes": 2321724,
-      "pages": 26,
+      "seconds": 2.1994080930016935,
+      "bytes": 6022375,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.6746519440002885,
-      "bytes": 10811116,
-      "pages": 42,
+      "seconds": 1.5225848519985448,
+      "bytes": 6009508,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.732792734999748,
-      "bytes": 6311900,
-      "pages": 36,
+      "seconds": 1.197222058999614,
+      "bytes": 6186205,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1773680320002313,
-      "bytes": 2462043,
-      "pages": 33,
+      "seconds": 1.0562214180026785,
+      "bytes": 2441509,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.2794374899985996,
-      "bytes": 6045695,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.3156736879973323,
+      "bytes": 6047068,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.710491572997853,
-      "bytes": 17528313,
-      "pages": 40,
-      "shards": 5
+      "seconds": 1.5381603519999771,
+      "bytes": 16328441,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 1.008943750999606,
-      "bytes": 1209380,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.2096843409999565,
+      "bytes": 2710723,
+      "pages": 38,
+      "shards": 2
     },
     {
-      "seconds": 1.5094498580001527,
-      "bytes": 5827665,
-      "pages": 28,
+      "seconds": 1.2958958110029926,
+      "bytes": 5955386,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0169636540013016,
-      "bytes": 2368292,
-      "pages": 29,
+      "seconds": 1.584322863000125,
+      "bytes": 6087710,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3777095789992018,
-      "bytes": 11141641,
-      "pages": 38,
+      "seconds": 1.2062028339969402,
+      "bytes": 5756694,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 3.434384232998127,
-      "bytes": 5971898,
+      "seconds": 1.087463227002445,
+      "bytes": 6477149,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1304158860002644,
-      "bytes": 8230136,
+      "seconds": 1.7014945069968235,
+      "bytes": 13448331,
       "pages": 37,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.2011196589992323,
-      "bytes": 6075797,
-      "pages": 37,
+      "seconds": 1.4291834129980998,
+      "bytes": 4870018,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.276424440999108,
-      "bytes": 6070440,
-      "pages": 25,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2386367050021363,
-      "bytes": 6009209,
-      "pages": 36,
+      "seconds": 1.6493652240023948,
+      "bytes": 6440305,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1234037470021576,
-      "bytes": 6172938,
-      "pages": 37,
+      "seconds": 1.0658796659990912,
+      "bytes": 2378111,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.8850959770024929,
-      "bytes": 2438669,
-      "pages": 31,
+      "seconds": 1.2226023729999724,
+      "bytes": 6024557,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.6140179690009973,
-      "bytes": 15868305,
-      "pages": 31,
+      "seconds": 2.320106137998664,
+      "bytes": 11400278,
+      "pages": 37,
       "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,
+      "seconds": 1.6032572549993347,
+      "bytes": 5165320,
+      "pages": 25,
+      "shards": 2
+    },
+    {
+      "seconds": 1.8941422179996152,
+      "bytes": 6019433,
+      "pages": 39,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3825668060017051,
+      "bytes": 13491032,
       "pages": 40,
-      "shards": 3
+      "shards": 5
     },
     {
-      "seconds": 2.2792270009995264,
-      "bytes": 8233418,
-      "pages": 31,
+      "seconds": 0.9105340399983106,
+      "bytes": 1249369,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.295263830001204,
+      "bytes": 5736433,
+      "pages": 23,
+      "shards": 2
+    },
+    {
+      "seconds": 1.218638172998908,
+      "bytes": 7112834,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 1.1861695589977899,
-      "bytes": 6023568,
+      "seconds": 1.2346669860016846,
+      "bytes": 6391793,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0451825370000734,
-      "bytes": 965218,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.537398492997454,
+      "bytes": 7291338,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.2396269109995046,
-      "bytes": 7721145,
-      "pages": 37,
+      "seconds": 1.5484525030005898,
+      "bytes": 10816659,
+      "pages": 38,
       "shards": 3
+    },
+    {
+      "seconds": 2.108060000999103,
+      "bytes": 5732876,
+      "pages": 29,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4
-size 721773631
+oid sha256:8ba9bbe2eb54c264d2b9324ec47dc07ced3c51c8d254214772fea4aa3594fd98
+size 721768447

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,