Qwen90M step 355000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 350000,
+  "step": 355000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "f5426f370d1251d7bba5de89b56b33e67df71399f079e2eeb4e3700f38dc4b31",
+    "model.safetensors": "e07e1e6496a6b48cf11640c7168b06232c8e131cc65408e66ebbe90be5fcabf0",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "b2c2d3d080843d366ffbe0938a83b313ceff09a34962f3581968203e2d4b079d",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "27deb3f82952932a7dca09467002556e1f8621d42020878ad6f833309e2c4d60",
+    "training_config.json": "a9be8fd9bf1c8d33aae0c56259fc899809df0d3ff60285ef0a4c2ea73471632c",
     "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a",
-    "shard_scheduler.json": "fad92bf0a8e73e7059f462c17088165edab733c4deb3ec42d95e84132126c970",
-    "coverage.json": "c9baf1742338a56c30b328003396bc082f4bfbced4f8bd6b990d6a8a6cef1d28",
+    "shard_scheduler.json": "a586d4c8276b86561726c239519a16e3d56185f24feafbd8f72991b1ac951402",
+    "coverage.json": "c48347a79685e76cda2c7abe60994cb73ba2a2da3ecf8adff831efd7fc00c98b",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 350000,
+  "step": 355000,
   "stats": {
-    "climbmix/blocks": 12600000,
-    "climbmix/targets": 21063108714,
-    "climbmix/input_tokens": 21064443608,
-    "ultra_qa/blocks": 4725000,
-    "ultra_qa/targets": 7708127333,
-    "ultra_qa/input_tokens": 7708142241,
-    "cortex/blocks": 4725000,
-    "cortex/targets": 179755192,
-    "cortex/input_tokens": 7511179034,
-    "ultra_style/blocks": 4725000,
-    "ultra_style/targets": 7473844592,
-    "ultra_style/input_tokens": 7474236913,
-    "openmath/blocks": 3150000,
-    "openmath/targets": 4867680878,
-    "openmath/input_tokens": 4869303434,
-    "rewrite6/blocks": 1575000,
-    "rewrite6/targets": 2581925316,
-    "rewrite6/input_tokens": 2584756299
+    "climbmix/blocks": 12780000,
+    "climbmix/targets": 21363361651,
+    "climbmix/input_tokens": 21364715661,
+    "ultra_qa/blocks": 4792500,
+    "ultra_qa/targets": 7818219856,
+    "ultra_qa/input_tokens": 7818234963,
+    "cortex/blocks": 4792500,
+    "cortex/targets": 182318776,
+    "cortex/input_tokens": 7618501449,
+    "ultra_style/blocks": 4792500,
+    "ultra_style/targets": 7580344733,
+    "ultra_style/input_tokens": 7580742835,
+    "openmath/blocks": 3195000,
+    "openmath/targets": 4937114048,
+    "openmath/input_tokens": 4938759767,
+    "rewrite6/blocks": 1597500,
+    "rewrite6/targets": 2618935783,
+    "rewrite6/input_tokens": 2621807196
   },
   "sources": {
     "climbmix": {
-      "page": 1725,
-      "row": 3105,
+      "page": 1750,
+      "row": 2459,
       "start": {
-        "file": 83,
+        "file": 84,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 41,
+          "row_group": 58,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 7068705
+      "documents": 7170459
     },
     "rewrite6": {
-      "page": 31,
-      "row": 3505,
+      "page": 45,
+      "row": 3686,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 126976
+          "row": 184320
         }
       },
       "epoch": 19,
       "offset": 0,
-      "documents": 4033157
+      "documents": 4090682
     },
     "openmath": {
-      "page": 1078,
-      "row": 2678,
+      "page": 1094,
+      "row": 149,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 48,
-          "row": 988
+          "row_group": 114,
+          "row": 524
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 4418166
+      "documents": 4481173
     },
     "ultra_qa": {
-      "page": 466,
-      "row": 1192,
+      "page": 472,
+      "row": 3902,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 26875
+          "row": 51451
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1909928
+      "documents": 1937214
     },
     "ultra_style": {
-      "page": 711,
-      "row": 3707,
+      "page": 722,
+      "row": 802,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 2,
-          "row": 2093
+          "row": 47149
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2915963
+      "offset": 0,
+      "documents": 2958114
     },
     "cortex": {
-      "page": 91642,
-      "row": 117,
+      "page": 92951,
+      "row": 78,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 11730293
+      "documents": 11897806
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 350000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f5426f370d1251d7bba5de89b56b33e67df71399f079e2eeb4e3700f38dc4b31", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "b2c2d3d080843d366ffbe0938a83b313ceff09a34962f3581968203e2d4b079d", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "fad92bf0a8e73e7059f462c17088165edab733c4deb3ec42d95e84132126c970", "coverage.json": "c9baf1742338a56c30b328003396bc082f4bfbced4f8bd6b990d6a8a6cef1d28", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000350000"}
+{"step": 355000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e07e1e6496a6b48cf11640c7168b06232c8e131cc65408e66ebbe90be5fcabf0", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "27deb3f82952932a7dca09467002556e1f8621d42020878ad6f833309e2c4d60", "training_config.json": "a9be8fd9bf1c8d33aae0c56259fc899809df0d3ff60285ef0a4c2ea73471632c", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "a586d4c8276b86561726c239519a16e3d56185f24feafbd8f72991b1ac951402", "coverage.json": "c48347a79685e76cda2c7abe60994cb73ba2a2da3ecf8adff831efd7fc00c98b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000355000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f5426f370d1251d7bba5de89b56b33e67df71399f079e2eeb4e3700f38dc4b31
+oid sha256:e07e1e6496a6b48cf11640c7168b06232c8e131cc65408e66ebbe90be5fcabf0
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 75583.06335444357,
-      "required_pages_per_second_upper_estimate": 0.006551348368040456
+      "required_token_positions_per_second": 73436.5976740253,
+      "required_pages_per_second_upper_estimate": 0.006365298163029331
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9447.882919305446,
-      "required_pages_per_second_upper_estimate": 0.003768901754948718
+      "required_token_positions_per_second": 9179.574709253162,
+      "required_pages_per_second_upper_estimate": 0.003661869598393634
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18895.765838610892,
-      "required_pages_per_second_upper_estimate": 0.004371202872642004
+      "required_token_positions_per_second": 18359.149418506324,
+      "required_pages_per_second_upper_estimate": 0.0042470661079772485
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28343.648757916337,
-      "required_pages_per_second_upper_estimate": 0.0017259795017248035
+      "required_token_positions_per_second": 27538.724127759484,
+      "required_pages_per_second_upper_estimate": 0.0016769638148614063
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28343.648757916337,
-      "required_pages_per_second_upper_estimate": 0.0021819725416258404
+      "required_token_positions_per_second": 27538.724127759484,
+      "required_pages_per_second_upper_estimate": 0.00212000721542238
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.042780299000014566,
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
-      "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28343.648757916337,
-      "required_pages_per_second_upper_estimate": 0.3519332574830989
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 27538.724127759484,
+      "required_pages_per_second_upper_estimate": 0.3419387874859938
     }
   ],
   "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": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 62,
+    "prefetch_batches": 60,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.11979209478852226,
-    "producer_p95_seconds": 0.38984344500204315,
-    "gpu_update_seconds": 0.975456626496543,
-    "producer_batches_per_second": 8.34779625287773,
-    "gpu_batches_per_second": 1.0251609070426915,
-    "producer_headroom": 8.142913171512593,
+    "producer_batch_seconds": 0.41315269255511566,
+    "producer_p95_seconds": 0.4113726069990662,
+    "gpu_update_seconds": 1.0039680804286195,
+    "producer_batches_per_second": 2.4204126416690297,
+    "gpu_batches_per_second": 0.9960476030005601,
+    "producer_headroom": 2.43001703370162,
     "sustainable": true,
-    "buffer_seconds": 60.47831084278566,
+    "buffer_seconds": 60.23808482571717,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 379734.8500496105,
-    "upload_bytes_per_second": 4092962.040476732,
+    "generation_bytes_per_second": 324828.29635209485,
+    "upload_bytes_per_second": 5287585.875586291,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 81.48225904575573,
+    "estimated_full_shards_per_hour": 69.70059078142295,
     "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": 1.1287752040007035,
-      "bytes": 10564318,
-      "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0117927669998608,
-      "bytes": 847459,
-      "pages": 26,
+      "seconds": 0.9463060489979398,
+      "bytes": 743440,
+      "pages": 25,
       "shards": 1
     },
     {
-      "seconds": 1.3560152030004247,
-      "bytes": 13557643,
-      "pages": 39,
+      "seconds": 1.0638403249977273,
+      "bytes": 11964464,
+      "pages": 46,
       "shards": 4
     },
     {
-      "seconds": 1.124354845000198,
-      "bytes": 904261,
+      "seconds": 1.8820245999959297,
+      "bytes": 865291,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.0335092330024054,
-      "bytes": 5888838,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.3777121140010422,
+      "bytes": 7459243,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.2323054039989074,
-      "bytes": 11963716,
-      "pages": 40,
-      "shards": 4
+      "seconds": 2.28389226100262,
+      "bytes": 15733663,
+      "pages": 34,
+      "shards": 5
     },
     {
-      "seconds": 0.7193277299993497,
-      "bytes": 1174751,
-      "pages": 35,
+      "seconds": 0.7980627049983013,
+      "bytes": 1094775,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.1924678350005706,
-      "bytes": 7370627,
-      "pages": 30,
+      "seconds": 1.0217303430035827,
+      "bytes": 7986291,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.999219997000182,
-      "bytes": 6670501,
-      "pages": 27,
-      "shards": 2
+      "seconds": 0.8759991519982577,
+      "bytes": 805335,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.1155471289966954,
-      "bytes": 10846938,
-      "pages": 39,
+      "seconds": 1.0561498970055254,
+      "bytes": 10650758,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.1072990959983144,
-      "bytes": 7478866,
+      "seconds": 1.2883731410038308,
+      "bytes": 7659801,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9801313110001502,
-      "bytes": 949416,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2882471760021872,
-      "bytes": 6045719,
-      "pages": 36,
+      "seconds": 1.1231011929994565,
+      "bytes": 6897324,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.362017162002303,
-      "bytes": 11390848,
-      "pages": 38,
-      "shards": 5
+      "seconds": 1.0080412499955855,
+      "bytes": 7466769,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 0.9684943469983409,
-      "bytes": 6504817,
-      "pages": 23,
+      "seconds": 1.1737449919965002,
+      "bytes": 4963940,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.0333191259996966,
-      "bytes": 7473380,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8470484819990816,
-      "bytes": 1045811,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.0047582629995304,
+      "bytes": 6204635,
+      "pages": 42,
+      "shards": 2
     },
     {
-      "seconds": 1.2339288350012794,
-      "bytes": 10876745,
+      "seconds": 1.2032458940011566,
+      "bytes": 7462591,
       "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.0449587049988622,
-      "bytes": 2671135,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0467685680014256,
-      "bytes": 5999640,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.0979009180009598,
+      "bytes": 1059989,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.1004228650017467,
-      "bytes": 6536803,
-      "pages": 24,
-      "shards": 2
+      "seconds": 1.474424235006154,
+      "bytes": 14927789,
+      "pages": 21,
+      "shards": 4
     },
     {
-      "seconds": 1.3272996070008958,
-      "bytes": 12496214,
-      "pages": 42,
-      "shards": 4
+      "seconds": 1.0885418239995488,
+      "bytes": 8199005,
+      "pages": 47,
+      "shards": 3
     },
     {
-      "seconds": 1.0267054780015314,
-      "bytes": 5908271,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.9340340310009196,
+      "bytes": 958542,
+      "pages": 30,
+      "shards": 1
     },
     {
-      "seconds": 0.8663640420018055,
-      "bytes": 2740775,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.0266037770052208,
+      "bytes": 7314912,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 0.9017275879996305,
-      "bytes": 5703769,
-      "pages": 28,
+      "seconds": 1.1214634670031955,
+      "bytes": 5616787,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2448231849994045,
-      "bytes": 5850047,
-      "pages": 33,
+      "seconds": 1.0272974420004175,
+      "bytes": 6127988,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1455546180004603,
-      "bytes": 12663721,
+      "seconds": 1.1608434069930809,
+      "bytes": 13413721,
       "pages": 38,
       "shards": 4
     },
     {
-      "seconds": 1.2500024359978852,
-      "bytes": 6221044,
+      "seconds": 0.8657637560027069,
+      "bytes": 989817,
       "pages": 27,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 0.9838440579987946,
-      "bytes": 1078556,
-      "pages": 35,
+      "seconds": 0.8787106930030859,
+      "bytes": 615085,
+      "pages": 17,
       "shards": 1
     },
     {
-      "seconds": 1.215074095998716,
-      "bytes": 7808319,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.1383813439970254,
+      "bytes": 11307353,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.0071898829992278,
-      "bytes": 5466457,
+      "seconds": 0.9193243289992097,
+      "bytes": 5971902,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1624265700011165,
-      "bytes": 7895099,
-      "pages": 39,
+      "seconds": 0.7975082029952318,
+      "bytes": 1112881,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.220746667997446,
+      "bytes": 11991108,
+      "pages": 31,
+      "shards": 4
+    },
+    {
+      "seconds": 7.638744584997767,
+      "bytes": 11951922,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.032417123002233,
-      "bytes": 5759947,
-      "pages": 28,
+      "seconds": 0.8686913920028019,
+      "bytes": 2634908,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9372151059978933,
-      "bytes": 7005197,
-      "pages": 37,
+      "seconds": 1.1666307280029287,
+      "bytes": 5700254,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 34.0496005649984,
-      "bytes": 11526687,
-      "pages": 38,
-      "shards": 4
+      "seconds": 0.8491491380045773,
+      "bytes": 1087747,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.0691167909972137,
-      "bytes": 6134154,
+      "seconds": 1.1218249969970202,
+      "bytes": 11930213,
       "pages": 36,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.9025495240020973,
-      "bytes": 2606249,
-      "pages": 37,
+      "seconds": 1.2040022809960647,
+      "bytes": 6093346,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.0264649749988166,
-      "bytes": 10585305,
+      "seconds": 0.9920474090031348,
+      "bytes": 6801926,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.7573466569992888,
-      "bytes": 999325,
-      "pages": 29,
-      "shards": 1
+      "seconds": 1.4719860830009566,
+      "bytes": 7846458,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.4190698960010195,
-      "bytes": 13694673,
-      "pages": 40,
-      "shards": 4
+      "seconds": 0.9881938749967958,
+      "bytes": 4661656,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 0.8739250739999989,
-      "bytes": 5944089,
+      "seconds": 1.2476324340022984,
+      "bytes": 7151046,
       "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9890799559943844,
+      "bytes": 6390118,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9463060489979398,
-      "bytes": 743440,
-      "pages": 25,
+      "seconds": 1.1548123519969522,
+      "bytes": 1183539,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.0638403249977273,
-      "bytes": 11964464,
-      "pages": 46,
-      "shards": 4
+      "seconds": 1.2286420309974346,
+      "bytes": 17809213,
+      "pages": 32,
+      "shards": 5
     },
     {
-      "seconds": 1.8820245999959297,
-      "bytes": 865291,
-      "pages": 27,
+      "seconds": 0.9948159990017302,
+      "bytes": 1198699,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.3777121140010422,
-      "bytes": 7459243,
-      "pages": 37,
+      "seconds": 1.0069344939984148,
+      "bytes": 6086180,
+      "pages": 31,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1158472850002,
+      "bytes": 7287693,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 2.28389226100262,
-      "bytes": 15733663,
+      "seconds": 1.026877347001573,
+      "bytes": 5561907,
       "pages": 34,
-      "shards": 5
+      "shards": 2
     },
     {
-      "seconds": 0.7980627049983013,
-      "bytes": 1094775,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.067893715997343,
+      "bytes": 7222726,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.0217303430035827,
-      "bytes": 7986291,
+      "seconds": 1.3662753429962322,
+      "bytes": 11997986,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.8759991519982577,
-      "bytes": 805335,
+      "seconds": 0.77281211099762,
+      "bytes": 859716,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.0561498970055254,
-      "bytes": 10650758,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2883731410038308,
-      "bytes": 7659801,
+      "seconds": 1.1721114309984841,
+      "bytes": 11272182,
       "pages": 38,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.1231011929994565,
-      "bytes": 6897324,
+      "seconds": 0.8626750620023813,
+      "bytes": 1117608,
       "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0080412499955855,
-      "bytes": 7466769,
-      "pages": 29,
-      "shards": 3
+      "shards": 1
     },
     {
-      "seconds": 1.1737449919965002,
-      "bytes": 4963940,
-      "pages": 26,
+      "seconds": 0.9938163240003632,
+      "bytes": 5819862,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0047582629995304,
-      "bytes": 6204635,
-      "pages": 42,
-      "shards": 2
+      "seconds": 1.1086062760005007,
+      "bytes": 7618244,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.2032458940011566,
-      "bytes": 7462591,
+      "seconds": 1.1348141839989694,
+      "bytes": 11320691,
       "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.0979009180009598,
-      "bytes": 1059989,
-      "pages": 33,
-      "shards": 1
+      "seconds": 0.981941227997595,
+      "bytes": 7228716,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.474424235006154,
-      "bytes": 14927789,
-      "pages": 21,
-      "shards": 4
+      "seconds": 0.914243400999112,
+      "bytes": 1165448,
+      "pages": 36,
+      "shards": 1
     },
     {
-      "seconds": 1.0885418239995488,
-      "bytes": 8199005,
-      "pages": 47,
-      "shards": 3
+      "seconds": 1.0446478919984656,
+      "bytes": 5999613,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9340340310009196,
-      "bytes": 958542,
+      "seconds": 1.4630355209956178,
+      "bytes": 11583159,
       "pages": 30,
-      "shards": 1
+      "shards": 4
     },
     {
-      "seconds": 1.0266037770052208,
-      "bytes": 7314912,
+      "seconds": 1.1508738010015804,
+      "bytes": 1166291,
       "pages": 35,
-      "shards": 3
+      "shards": 1
     },
     {
-      "seconds": 1.1214634670031955,
-      "bytes": 5616787,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.4934369809998316,
+      "bytes": 13370474,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.0272974420004175,
-      "bytes": 6127988,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.8853696480000508,
+      "bytes": 1220951,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.1608434069930809,
-      "bytes": 13413721,
-      "pages": 38,
-      "shards": 4
+      "seconds": 2.1309150849992875,
+      "bytes": 9524825,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1543053290006355,
+      "bytes": 7653463,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0381377000012435,
+      "bytes": 839375,
+      "pages": 26,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b2c2d3d080843d366ffbe0938a83b313ceff09a34962f3581968203e2d4b079d
-size 721769855
+oid sha256:27deb3f82952932a7dca09467002556e1f8621d42020878ad6f833309e2c4d60
+size 721770687

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 62,
-  "prefetch_low_water": 31,
+  "prefetch_batches": 60,
+  "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,