Qwen90M step 625000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 620000,
+  "step": 625000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "5e1c8846d93cc736a471c45613d78007b35b8e94c1285a2904ac37057a5c3356",
+    "model.safetensors": "e79b91ffdb8a4225e499aae34fbdee578329f2c0cbb6e3ff0bd680339364d489",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "9933156547ab18b657069aa60eb52e3b669e25faa068a6484883abec26f0f1a7",
-    "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
+    "trainer.pt": "08c9cb9f6f17586214603f83e8a6ed392e1f2c67500da7bfa88fe516501b5bdf",
+    "training_config.json": "99d98cec9c0814e63e7b7a5836277d5c336d3c2fb33fa69da403f3610c902c99",
     "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e",
-    "shard_scheduler.json": "b2d31325d348832dedfe0eeb8c5de7f5606e001abaa7474b129c45c2b3901479",
-    "coverage.json": "bb539a2055773d2d68de8915bf0ebe63f0b3e8e118e9bf2703f2ab7066ea0769",
+    "shard_scheduler.json": "e43f0cf9936e7a3794ba03cd0134b307fbf0346bb197cf8898bd22441a49a669",
+    "coverage.json": "06fc3d77f8d5cf5a9663605fb6f19290a742c2d87610dbe416069837444aa522",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 620000,
+  "step": 625000,
   "stats": {
-    "climbmix/blocks": 22320000,
-    "climbmix/targets": 37312355817,
-    "climbmix/input_tokens": 37314720200,
-    "ultra_qa/blocks": 8370000,
-    "ultra_qa/targets": 13655080912,
-    "ultra_qa/input_tokens": 13655107298,
-    "cortex/blocks": 8370000,
-    "cortex/targets": 318257872,
-    "cortex/input_tokens": 13304792682,
-    "ultra_style/blocks": 8370000,
-    "ultra_style/targets": 13238688750,
-    "ultra_style/input_tokens": 13239383690,
-    "openmath/blocks": 5580000,
-    "openmath/targets": 8622121150,
-    "openmath/input_tokens": 8624993272,
-    "rewrite6/blocks": 2790000,
-    "rewrite6/targets": 4573810627,
-    "rewrite6/input_tokens": 4578825235
+    "climbmix/blocks": 22500000,
+    "climbmix/targets": 37613482229,
+    "climbmix/input_tokens": 37615865584,
+    "ultra_qa/blocks": 8437500,
+    "ultra_qa/targets": 13765166892,
+    "ultra_qa/input_tokens": 13765193488,
+    "cortex/blocks": 8437500,
+    "cortex/targets": 320825520,
+    "cortex/input_tokens": 13412066305,
+    "ultra_style/blocks": 8437500,
+    "ultra_style/targets": 13345548414,
+    "ultra_style/input_tokens": 13346249095,
+    "openmath/blocks": 5625000,
+    "openmath/targets": 8691637642,
+    "openmath/input_tokens": 8694532988,
+    "rewrite6/blocks": 2812500,
+    "rewrite6/targets": 4610720475,
+    "rewrite6/input_tokens": 4615775369
   },
   "sources": {
     "climbmix": {
-      "page": 3056,
-      "row": 2715,
+      "page": 3081,
+      "row": 1020,
       "start": {
-        "file": 147,
+        "file": 149,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 75,
+          "row_group": 12,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 12520091
+      "offset": 24576,
+      "documents": 12620796
     },
     "rewrite6": {
-      "page": 39,
-      "row": 610,
+      "page": 3,
+      "row": 45,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 159744
+          "row": 12288
         }
       },
-      "epoch": 34,
+      "epoch": 35,
       "offset": 0,
-      "documents": 7144090
+      "documents": 7201473
     },
     "openmath": {
-      "page": 1910,
-      "row": 970,
+      "page": 1925,
+      "row": 2658,
       "start": {
-        "file": 17,
+        "file": 18,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 400,
-          "row": 310
+          "row_group": 25,
+          "row": 100
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 7824330
+      "offset": 0,
+      "documents": 7887458
     },
     "ultra_qa": {
-      "page": 826,
-      "row": 27,
+      "page": 832,
+      "row": 2800,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 7499
+          "row": 32075
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 3383323
+      "documents": 3410672
     },
     "ultra_style": {
-      "page": 1260,
-      "row": 3840,
+      "page": 1271,
+      "row": 697,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 25036
+          "row": 70092
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 5164800
+      "offset": 2048,
+      "documents": 5206713
     },
     "cortex": {
-      "page": 162331,
-      "row": 125,
+      "page": 163641,
+      "row": 38,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 20778493
+      "documents": 20946086
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 620000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "5e1c8846d93cc736a471c45613d78007b35b8e94c1285a2904ac37057a5c3356", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9933156547ab18b657069aa60eb52e3b669e25faa068a6484883abec26f0f1a7", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "b2d31325d348832dedfe0eeb8c5de7f5606e001abaa7474b129c45c2b3901479", "coverage.json": "bb539a2055773d2d68de8915bf0ebe63f0b3e8e118e9bf2703f2ab7066ea0769", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000620000"}
+{"step": 625000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e79b91ffdb8a4225e499aae34fbdee578329f2c0cbb6e3ff0bd680339364d489", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "08c9cb9f6f17586214603f83e8a6ed392e1f2c67500da7bfa88fe516501b5bdf", "training_config.json": "99d98cec9c0814e63e7b7a5836277d5c336d3c2fb33fa69da403f3610c902c99", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "e43f0cf9936e7a3794ba03cd0134b307fbf0346bb197cf8898bd22441a49a669", "coverage.json": "06fc3d77f8d5cf5a9663605fb6f19290a742c2d87610dbe416069837444aa522", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000625000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e1c8846d93cc736a471c45613d78007b35b8e94c1285a2904ac37057a5c3356
+oid sha256:e79b91ffdb8a4225e499aae34fbdee578329f2c0cbb6e3ff0bd680339364d489
 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": 76942.96702956906,
-      "required_pages_per_second_upper_estimate": 0.005323092991386044
+      "required_token_positions_per_second": 74700.81649595029,
+      "required_pages_per_second_upper_estimate": 0.005167975814964294
     },
     {
       "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": 9617.870878696132,
-      "required_pages_per_second_upper_estimate": 0.0040268016929273925
+      "required_token_positions_per_second": 9337.602061993786,
+      "required_pages_per_second_upper_estimate": 0.003909458992052544
     },
     {
       "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": 19235.741757392265,
-      "required_pages_per_second_upper_estimate": 0.004006297727441522
+      "required_token_positions_per_second": 18675.204123987573,
+      "required_pages_per_second_upper_estimate": 0.003889552520774789
     },
     {
       "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": 28853.612636088397,
-      "required_pages_per_second_upper_estimate": 0.0017297300206179822
+      "required_token_positions_per_second": 28012.806185981357,
+      "required_pages_per_second_upper_estimate": 0.001679324957771178
     },
     {
       "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": 28853.612636088397,
-      "required_pages_per_second_upper_estimate": 0.0022432414688265327
+      "required_token_positions_per_second": 28012.806185981357,
+      "required_pages_per_second_upper_estimate": 0.002177872465647434
     },
     {
       "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": 28853.612636088397,
-      "required_pages_per_second_upper_estimate": 0.37010316230023216
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28012.806185981357,
+      "required_pages_per_second_upper_estimate": 0.35931819994588776
     }
   ],
   "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": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.18642643738270692,
-    "producer_p95_seconds": 0.23165708300075494,
-    "gpu_update_seconds": 0.9582162327021578,
-    "producer_batches_per_second": 5.364046076507607,
-    "gpu_batches_per_second": 1.0436057811085213,
-    "producer_headroom": 5.139916023471909,
+    "producer_batch_seconds": 0.43565610739844374,
+    "producer_p95_seconds": 0.2395410280005308,
+    "gpu_update_seconds": 0.9869771638171716,
+    "producer_batches_per_second": 2.2953884566696017,
+    "gpu_batches_per_second": 1.0131946681850896,
+    "producer_headroom": 2.265495988822438,
     "sustainable": true,
-    "buffer_seconds": 60.367622660235945,
+    "buffer_seconds": 60.205606992847464,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 81,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 480694.27939603303,
-    "upload_bytes_per_second": 5198415.335503838,
+    "generation_bytes_per_second": 322686.5127973047,
+    "upload_bytes_per_second": 5089683.248589828,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 103.14580236826652,
+    "estimated_full_shards_per_hour": 69.2410138887344,
     "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": 1.2262302330000239,
-      "bytes": 12002858,
+      "seconds": 1.2692650000008143,
+      "bytes": 13237909,
       "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 0.976691177000248,
-      "bytes": 6244802,
-      "pages": 36,
+      "seconds": 1.122997641999973,
+      "bytes": 6774100,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.5793915419999394,
-      "bytes": 781594,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4397608880008193,
-      "bytes": 13440043,
-      "pages": 39,
-      "shards": 4
-    },
-    {
-      "seconds": 1.215214173998902,
-      "bytes": 10529948,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3451746110004024,
-      "bytes": 2254637,
+      "seconds": 2.5239626219990896,
+      "bytes": 2432130,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1764542460005032,
-      "bytes": 6191157,
+      "seconds": 1.0105871389987442,
+      "bytes": 6118542,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0847412589992018,
-      "bytes": 1126127,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.3655256990005,
+      "bytes": 10029580,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.2304532270009076,
-      "bytes": 11321849,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.8411326719997305,
+      "bytes": 2272552,
+      "pages": 26,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4776745150011266,
+      "bytes": 12319008,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.4404141169998184,
-      "bytes": 12120623,
+      "seconds": 1.1878520110003592,
+      "bytes": 6586003,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.8672846369991021,
-      "bytes": 2691618,
-      "pages": 37,
+      "seconds": 4.034771163000187,
+      "bytes": 5791973,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 0.9980244999987917,
-      "bytes": 5788373,
-      "pages": 22,
+      "seconds": 1.1227300080008717,
+      "bytes": 6451625,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 3.3476856620000035,
-      "bytes": 5904691,
-      "pages": 43,
+      "seconds": 1.4268143489989598,
+      "bytes": 2671684,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2678156700003456,
-      "bytes": 7637773,
+      "seconds": 1.154910495999502,
+      "bytes": 10565027,
+      "pages": 23,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3083442860006471,
+      "bytes": 8430845,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1486175610007194,
-      "bytes": 5966085,
+      "seconds": 1.0249396090002847,
+      "bytes": 5941897,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2488192199998593,
+      "bytes": 5819043,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0020495059998211,
-      "bytes": 6673858,
-      "pages": 32,
+      "seconds": 0.903354562999084,
+      "bytes": 2640360,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1948771370007307,
-      "bytes": 12132833,
-      "pages": 38,
-      "shards": 4
+      "seconds": 1.1839327219986444,
+      "bytes": 10186345,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.2424271460004093,
-      "bytes": 1081337,
-      "pages": 34,
-      "shards": 1
+      "seconds": 3.6534681639986957,
+      "bytes": 12129367,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.3408090630000515,
-      "bytes": 7658166,
-      "pages": 39,
+      "seconds": 0.9705111450002732,
+      "bytes": 2305053,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3458375579994026,
+      "bytes": 11053239,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 0.8867853440006002,
-      "bytes": 5785267,
-      "pages": 27,
+      "seconds": 1.1655933139991248,
+      "bytes": 2619678,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.081288327000948,
-      "bytes": 5787210,
-      "pages": 36,
+      "seconds": 1.0188674239998363,
+      "bytes": 6227206,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1759565290012688,
-      "bytes": 13132586,
-      "pages": 31,
-      "shards": 4
+      "seconds": 0.9700920809991658,
+      "bytes": 6347868,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.004212905998429,
-      "bytes": 6157452,
-      "pages": 38,
+      "seconds": 1.00576553800056,
+      "bytes": 8295904,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.3965238540004066,
-      "bytes": 817171,
-      "pages": 26,
+      "seconds": 0.88294980000137,
+      "bytes": 116127,
+      "pages": 4,
       "shards": 1
     },
     {
-      "seconds": 1.115737056999933,
-      "bytes": 11496202,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.2832525270005135,
+      "bytes": 10171575,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.037322108000808,
-      "bytes": 1047345,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.4380360830000427,
+      "bytes": 7912633,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.0591198959991743,
-      "bytes": 7189815,
-      "pages": 31,
-      "shards": 3
+      "seconds": 0.8056872329998441,
+      "bytes": 861106,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.272207456000615,
-      "bytes": 11979936,
-      "pages": 41,
+      "seconds": 1.1610822420007025,
+      "bytes": 10322718,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1375752950007154,
-      "bytes": 4905532,
-      "pages": 33,
+      "seconds": 1.0416573500006052,
+      "bytes": 2564522,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.533432411000831,
-      "bytes": 7357007,
-      "pages": 30,
+      "seconds": 1.327567568001541,
+      "bytes": 11737942,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.1559075860004668,
-      "bytes": 1149523,
+      "seconds": 1.1044216970003617,
+      "bytes": 7366106,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.023783693999576,
+      "bytes": 5395680,
       "pages": 34,
-      "shards": 1
+      "shards": 2
     },
     {
-      "seconds": 1.0760069800016936,
-      "bytes": 5947734,
-      "pages": 28,
+      "seconds": 1.439521827000135,
+      "bytes": 6360408,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2793344790006813,
-      "bytes": 12370460,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.058846297000855,
+      "bytes": 7513643,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 0.8282303630003298,
-      "bytes": 647721,
-      "pages": 20,
+      "seconds": 0.8542627569986507,
+      "bytes": 1239253,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.3377682740010641,
-      "bytes": 13564102,
-      "pages": 48,
-      "shards": 4
+      "seconds": 1.5474993310017453,
+      "bytes": 17697706,
+      "pages": 35,
+      "shards": 5
     },
     {
-      "seconds": 1.359001171998898,
-      "bytes": 1213849,
-      "pages": 32,
+      "seconds": 0.7860386789980112,
+      "bytes": 1154184,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.1420412400002533,
-      "bytes": 10074433,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.9510616730003676,
+      "bytes": 6332480,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.157611688000543,
-      "bytes": 7621104,
-      "pages": 39,
+      "seconds": 1.1097759929980384,
+      "bytes": 7296844,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.0118182780006464,
-      "bytes": 802933,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.8535761809980613,
+      "bytes": 4986805,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.0449377190016094,
-      "bytes": 6509260,
-      "pages": 28,
+      "seconds": 1.1148564830000396,
+      "bytes": 6125180,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2692650000008143,
-      "bytes": 13237909,
-      "pages": 41,
+      "seconds": 1.2460403479999513,
+      "bytes": 13328563,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 1.122997641999973,
-      "bytes": 6774100,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.0476313660001324,
+      "bytes": 1088716,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 2.5239626219990896,
-      "bytes": 2432130,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.4290280249988427,
+      "bytes": 12076168,
+      "pages": 35,
+      "shards": 5
     },
     {
-      "seconds": 1.0105871389987442,
-      "bytes": 6118542,
+      "seconds": 1.0162399760010885,
+      "bytes": 1245993,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.0449379370002134,
+      "bytes": 6389820,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3655256990005,
-      "bytes": 10029580,
-      "pages": 38,
+      "seconds": 1.0748847499999101,
+      "bytes": 7299524,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.8411326719997305,
-      "bytes": 2272552,
-      "pages": 26,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4776745150011266,
-      "bytes": 12319008,
-      "pages": 37,
+      "seconds": 3.3975770770011877,
+      "bytes": 10836444,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1878520110003592,
-      "bytes": 6586003,
+      "seconds": 1.1278280619990255,
+      "bytes": 7877906,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 4.034771163000187,
-      "bytes": 5791973,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1227300080008717,
-      "bytes": 6451625,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4268143489989598,
-      "bytes": 2671684,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.147284323000349,
+      "bytes": 1234396,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.154910495999502,
-      "bytes": 10565027,
-      "pages": 23,
+      "seconds": 1.2498522079986287,
+      "bytes": 10921992,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.3083442860006471,
-      "bytes": 8430845,
+      "seconds": 1.2837640949983324,
+      "bytes": 7864429,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0249396090002847,
-      "bytes": 5941897,
-      "pages": 36,
-      "shards": 3
+      "seconds": 0.8372707349990378,
+      "bytes": 1184239,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.2488192199998593,
-      "bytes": 5819043,
+      "seconds": 1.1600354100010009,
+      "bytes": 11531600,
       "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 0.903354562999084,
-      "bytes": 2640360,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1839327219986444,
-      "bytes": 10186345,
-      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 3.6534681639986957,
-      "bytes": 12129367,
-      "pages": 40,
+      "seconds": 1.4267704899975797,
+      "bytes": 7513990,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9705111450002732,
-      "bytes": 2305053,
-      "pages": 28,
+      "seconds": 1.0001321140007349,
+      "bytes": 5054764,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.3458375579994026,
-      "bytes": 11053239,
-      "pages": 36,
+      "seconds": 1.5159087600004568,
+      "bytes": 7203777,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.1655933139991248,
-      "bytes": 2619678,
+      "seconds": 1.1342419799984782,
+      "bytes": 1205083,
       "pages": 35,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.0188674239998363,
-      "bytes": 6227206,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2808730480028316,
+      "bytes": 10383198,
+      "pages": 27,
+      "shards": 3
     },
     {
-      "seconds": 0.9700920809991658,
-      "bytes": 6347868,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.2442956449995108,
+      "bytes": 13618701,
+      "pages": 45,
+      "shards": 4
     },
     {
-      "seconds": 1.00576553800056,
-      "bytes": 8295904,
-      "pages": 33,
+      "seconds": 0.9636815740013844,
+      "bytes": 1058213,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1714537059997383,
+      "bytes": 7311656,
+      "pages": 31,
       "shards": 3
+    },
+    {
+      "seconds": 2.4602762009999424,
+      "bytes": 844768,
+      "pages": 28,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9933156547ab18b657069aa60eb52e3b669e25faa068a6484883abec26f0f1a7
-size 721780991
+oid sha256:08c9cb9f6f17586214603f83e8a6ed392e1f2c67500da7bfa88fe516501b5bdf
+size 721760319

training_config.json

@@ -53,8 +53,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
-  "prefetch_low_water": 31,
+  "prefetch_batches": 61,
+  "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,