Qwen90M step 155000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 150000,
+  "step": 155000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b",
+    "model.safetensors": "118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4",
-    "shard_scheduler.json": "34c9075a0f66a3d08c183399d506bb22fcf18c184ad147564967e4708a38ba3f",
-    "coverage.json": "a2f785366449384f55526192ff2e3551a8f132be1f93c9467f8d4fa84286825c",
+    "shard_scheduler.json": "81a7ca71b967fe09918b8d8fc0d0910018e4a0b46e1c1831d67cb3e3d678dfdb",
+    "coverage.json": "66656093e66bc4071d1aa3216cfbf69575a2e82a31d04761ddf384991f4d7c54",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 150000,
+  "step": 155000,
   "stats": {
-    "climbmix/blocks": 5400000,
-    "climbmix/targets": 9024151403,
-    "climbmix/input_tokens": 9024723849,
-    "ultra_qa/blocks": 2025000,
-    "ultra_qa/targets": 3303216837,
-    "ultra_qa/input_tokens": 3303223204,
-    "cortex/blocks": 2025000,
-    "cortex/targets": 77013811,
-    "cortex/input_tokens": 3218659538,
-    "ultra_style/blocks": 2025000,
-    "ultra_style/targets": 3203109017,
-    "ultra_style/input_tokens": 3203277056,
-    "openmath/blocks": 1350000,
-    "openmath/targets": 2085884301,
-    "openmath/input_tokens": 2086579643,
-    "rewrite6/blocks": 675000,
-    "rewrite6/targets": 1106511308,
-    "rewrite6/input_tokens": 1107724623
+    "climbmix/blocks": 5580000,
+    "climbmix/targets": 9325510201,
+    "climbmix/input_tokens": 9326101685,
+    "ultra_qa/blocks": 2092500,
+    "ultra_qa/targets": 3413363253,
+    "ultra_qa/input_tokens": 3413369829,
+    "cortex/blocks": 2092500,
+    "cortex/targets": 79585819,
+    "cortex/input_tokens": 3325972346,
+    "ultra_style/blocks": 2092500,
+    "ultra_style/targets": 3309559754,
+    "ultra_style/input_tokens": 3309733622,
+    "openmath/blocks": 1395000,
+    "openmath/targets": 2155399566,
+    "openmath/input_tokens": 2156118041,
+    "rewrite6/blocks": 697500,
+    "rewrite6/targets": 1143387147,
+    "rewrite6/input_tokens": 1144640945
   },
   "sources": {
     "climbmix": {
-      "page": 740,
-      "row": 1687,
+      "page": 764,
+      "row": 3782,
       "start": {
-        "file": 35,
+        "file": 36,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 60,
+          "row_group": 74,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 3032727
+      "offset": 4096,
+      "documents": 3133126
     },
     "rewrite6": {
-      "page": 20,
-      "row": 3357,
+      "page": 34,
+      "row": 3725,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 81920
+          "row": 139264
         }
       },
       "epoch": 8,
       "offset": 0,
-      "documents": 1728509
+      "documents": 1786221
     },
     "openmath": {
-      "page": 462,
-      "row": 1399,
+      "page": 477,
+      "row": 2923,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 145,
-          "row": 752
+          "row_group": 207,
+          "row": 192
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1893751
+      "offset": 0,
+      "documents": 1956715
     },
     "ultra_qa": {
-      "page": 199,
-      "row": 3551,
+      "page": 206,
+      "row": 2172,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
-          "row": 37088
+          "row_group": 5,
+          "row": 1242
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 818655
+      "documents": 845948
     },
     "ultra_style": {
-      "page": 305,
-      "row": 256,
+      "page": 315,
+      "row": 1608,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 6,
-          "row": 54213
+          "row_group": 7,
+          "row": 9777
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1249536
+      "documents": 1291848
     },
     "cortex": {
-      "page": 39271,
-      "row": 122,
+      "page": 40579,
+      "row": 14,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 5026810
+      "documents": 5194126
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 150000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4", "shard_scheduler.json": "34c9075a0f66a3d08c183399d506bb22fcf18c184ad147564967e4708a38ba3f", "coverage.json": "a2f785366449384f55526192ff2e3551a8f132be1f93c9467f8d4fa84286825c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000150000"}
+{"step": 155000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4", "shard_scheduler.json": "81a7ca71b967fe09918b8d8fc0d0910018e4a0b46e1c1831d67cb3e3d678dfdb", "coverage.json": "66656093e66bc4071d1aa3216cfbf69575a2e82a31d04761ddf384991f4d7c54", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000155000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b
+oid sha256:118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c
 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": 75368.18777639217,
-      "required_pages_per_second_upper_estimate": 0.006532723497532133
+      "required_token_positions_per_second": 74435.92703712951,
+      "required_pages_per_second_upper_estimate": 0.006451917499446089
     },
     {
       "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": 9421.02347204902,
-      "required_pages_per_second_upper_estimate": 0.0037581871198536063
+      "required_token_positions_per_second": 9304.49087964119,
+      "required_pages_per_second_upper_estimate": 0.0037117005264246006
     },
     {
       "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": 18842.04694409804,
-      "required_pages_per_second_upper_estimate": 0.004358775951816709
+      "required_token_positions_per_second": 18608.98175928238,
+      "required_pages_per_second_upper_estimate": 0.0043048604231167645
     },
     {
       "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": 28263.070416147064,
-      "required_pages_per_second_upper_estimate": 0.0017210727034729413
+      "required_token_positions_per_second": 27913.47263892357,
+      "required_pages_per_second_upper_estimate": 0.0016997840330378133
     },
     {
       "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": 28263.070416147064,
-      "required_pages_per_second_upper_estimate": 0.0021757693978213155
+      "required_token_positions_per_second": 27913.47263892357,
+      "required_pages_per_second_upper_estimate": 0.002148856393182065
     },
     {
       "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": 28263.070416147064,
-      "required_pages_per_second_upper_estimate": 0.3509327441566865
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 27913.47263892357,
+      "required_pages_per_second_upper_estimate": 0.3465919097920654
     }
   ],
   "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": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.1606457544066302,
-    "producer_p95_seconds": 0.3911977770039812,
-    "gpu_update_seconds": 0.9782376646595458,
-    "producer_batches_per_second": 6.2248766155922,
-    "gpu_batches_per_second": 1.0222464704914302,
-    "producer_headroom": 6.089408763230731,
+    "producer_batch_seconds": 0.1047992101799764,
+    "producer_p95_seconds": 0.4328458870004397,
+    "gpu_update_seconds": 0.9904894442064731,
+    "producer_batches_per_second": 9.54205664606303,
+    "gpu_batches_per_second": 1.0096018749610667,
+    "producer_headroom": 9.451306383945653,
     "sustainable": true,
-    "buffer_seconds": 60.65073520889184,
+    "buffer_seconds": 60.41985609659486,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 362750.0021895322,
-    "upload_bytes_per_second": 5470118.08714139,
+    "generation_bytes_per_second": 341840.10677463334,
+    "upload_bytes_per_second": 5861021.62228704,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 77.83770608200525,
+    "estimated_full_shards_per_hour": 73.35092928342104,
     "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": 1.7747774790004769,
-      "bytes": 7374031,
-      "pages": 33,
+      "seconds": 1.2166951500003051,
+      "bytes": 7600114,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9477112800013856,
-      "bytes": 5792070,
-      "pages": 34,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0662493220006581,
-      "bytes": 5942385,
+      "seconds": 1.056514903997595,
+      "bytes": 5902748,
       "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.245676413000183,
-      "bytes": 7693974,
-      "pages": 38,
+      "seconds": 1.0406464490006329,
+      "bytes": 6219293,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.0202759780004271,
-      "bytes": 6615812,
-      "pages": 28,
+      "seconds": 1.1162217770033749,
+      "bytes": 6495200,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4830753700007335,
-      "bytes": 9974834,
-      "pages": 35,
+      "seconds": 1.1027901979978196,
+      "bytes": 11991791,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.272736505001376,
-      "bytes": 7611348,
-      "pages": 38,
+      "seconds": 1.5872819810028886,
+      "bytes": 6214149,
+      "pages": 24,
       "shards": 3
     },
     {
-      "seconds": 0.8344998959983059,
-      "bytes": 1093697,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.0351565999953891,
+      "bytes": 6216009,
+      "pages": 41,
+      "shards": 2
     },
     {
-      "seconds": 1.3662152600008994,
-      "bytes": 7153184,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.4164216759963892,
+      "bytes": 2450413,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.6717624239972793,
-      "bytes": 5422811,
-      "pages": 35,
+      "seconds": 1.0167613629964762,
+      "bytes": 6214181,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.1521110989997396,
-      "bytes": 12146138,
-      "pages": 38,
+      "seconds": 1.3737143940015812,
+      "bytes": 10463664,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.1019259579989011,
-      "bytes": 7338055,
-      "pages": 30,
+      "seconds": 0.8019685460021719,
+      "bytes": 2186825,
+      "pages": 23,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2952469239971833,
+      "bytes": 12034549,
+      "pages": 44,
       "shards": 3
     },
     {
-      "seconds": 0.7986074580003333,
-      "bytes": 1164850,
-      "pages": 35,
-      "shards": 1
+      "seconds": 0.8924442770003225,
+      "bytes": 6155396,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.6704007100015588,
-      "bytes": 10030549,
+      "seconds": 1.1965257189949625,
+      "bytes": 6621394,
       "pages": 31,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 2.951283398000669,
-      "bytes": 2617533,
-      "pages": 35,
+      "seconds": 1.1455287000062526,
+      "bytes": 6484150,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1606040469996515,
-      "bytes": 6236824,
-      "pages": 36,
+      "seconds": 0.8978742559993407,
+      "bytes": 2461928,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 4.655135680000967,
-      "bytes": 7374926,
-      "pages": 30,
-      "shards": 3
+      "seconds": 0.9092525360028958,
+      "bytes": 6015320,
+      "pages": 26,
+      "shards": 2
     },
     {
-      "seconds": 1.311550709000585,
-      "bytes": 11722471,
+      "seconds": 1.3533267960010562,
+      "bytes": 11703451,
       "pages": 42,
       "shards": 3
     },
     {
-      "seconds": 0.7739134580006066,
-      "bytes": 5816616,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.117633730002126,
-      "bytes": 7565346,
-      "pages": 38,
+      "seconds": 1.2227107450016774,
+      "bytes": 7523214,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 0.8082868439996673,
-      "bytes": 1195987,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.157742789000622,
-      "bytes": 11040923,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0604035779979313,
+      "bytes": 6119324,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.8454459890017461,
-      "bytes": 981466,
-      "pages": 29,
+      "seconds": 1.0826924209977733,
+      "bytes": 902486,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.0536112129993853,
-      "bytes": 11943546,
-      "pages": 39,
+      "seconds": 1.1300200109981233,
+      "bytes": 7391590,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0575292020002962,
-      "bytes": 2082389,
-      "pages": 19,
+      "seconds": 0.8869208329997491,
+      "bytes": 4199574,
+      "pages": 23,
       "shards": 2
     },
     {
-      "seconds": 1.061086557998351,
-      "bytes": 10157591,
+      "seconds": 1.295291658003407,
+      "bytes": 11877354,
       "pages": 32,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 1.351508649000607,
-      "bytes": 6405900,
-      "pages": 37,
+      "seconds": 1.0650672470001155,
+      "bytes": 2271023,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 0.9840362190006999,
-      "bytes": 2390157,
-      "pages": 28,
+      "seconds": 1.0174418120004702,
+      "bytes": 5974488,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.0104034229989338,
-      "bytes": 6117004,
-      "pages": 36,
+      "seconds": 1.083592351002153,
+      "bytes": 5253625,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.275026989998878,
-      "bytes": 16016482,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.5616185160033638,
+      "bytes": 7297224,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 0.8882673849984712,
-      "bytes": 2492218,
-      "pages": 31,
+      "seconds": 1.1017810449993704,
+      "bytes": 6135514,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2180680759993265,
-      "bytes": 6022567,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.7622847379971063,
+      "bytes": 678366,
+      "pages": 20,
+      "shards": 1
     },
     {
-      "seconds": 1.038481719999254,
-      "bytes": 2380066,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.3286253929982195,
+      "bytes": 17203756,
+      "pages": 37,
+      "shards": 5
     },
     {
-      "seconds": 1.0937937710004917,
-      "bytes": 10154198,
-      "pages": 42,
+      "seconds": 0.8924150489983731,
+      "bytes": 1133711,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.4622295160006615,
+      "bytes": 7475361,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.047995336000895,
-      "bytes": 5924124,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.0509287640015827,
+      "bytes": 1010925,
+      "pages": 30,
+      "shards": 1
     },
     {
-      "seconds": 1.1032993179978803,
-      "bytes": 8345826,
-      "pages": 35,
+      "seconds": 0.9923871680002776,
+      "bytes": 11139165,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9737631519965362,
-      "bytes": 6188431,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.103713971999241,
+      "bytes": 7399212,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.3008259939997515,
-      "bytes": 12099101,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.030579199003114,
+      "bytes": 6815769,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 0.8470429570006672,
-      "bytes": 932725,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.1106542109992006,
+      "bytes": 7819068,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.1173031580001407,
-      "bytes": 6247577,
-      "pages": 37,
+      "seconds": 0.8756177200048114,
+      "bytes": 4958523,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 0.892657821001194,
-      "bytes": 2615407,
-      "pages": 37,
+      "seconds": 0.9241842089977581,
+      "bytes": 5682376,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2370568289989023,
-      "bytes": 16078869,
-      "pages": 34,
-      "shards": 4
+      "seconds": 0.9258055570026045,
+      "bytes": 7457751,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 0.8470632479984488,
-      "bytes": 947755,
-      "pages": 29,
+      "seconds": 1.6118608709948603,
+      "bytes": 885264,
+      "pages": 28,
       "shards": 1
     },
     {
-      "seconds": 1.2166951500003051,
-      "bytes": 7600114,
-      "pages": 37,
+      "seconds": 1.013624004997837,
+      "bytes": 9878612,
+      "pages": 14,
       "shards": 3
     },
     {
-      "seconds": 1.056514903997595,
-      "bytes": 5902748,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.218867665003927,
+      "bytes": 8704871,
+      "pages": 46,
+      "shards": 3
     },
     {
-      "seconds": 1.0406464490006329,
-      "bytes": 6219293,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.0288945650027017,
+      "bytes": 5962293,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.1162217770033749,
-      "bytes": 6495200,
-      "pages": 37,
+      "seconds": 0.9867427620047238,
+      "bytes": 6108091,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.1027901979978196,
-      "bytes": 11991791,
-      "pages": 37,
+      "seconds": 0.9743170239962637,
+      "bytes": 2340987,
+      "pages": 25,
       "shards": 3
     },
     {
-      "seconds": 1.5872819810028886,
-      "bytes": 6214149,
-      "pages": 24,
+      "seconds": 1.4316346049963613,
+      "bytes": 10012608,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0351565999953891,
-      "bytes": 6216009,
-      "pages": 41,
-      "shards": 2
+      "seconds": 0.8118417049990967,
+      "bytes": 973375,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.4164216759963892,
-      "bytes": 2450413,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.5195651320027537,
+      "bytes": 13317753,
+      "pages": 32,
+      "shards": 4
     },
     {
-      "seconds": 1.0167613629964762,
-      "bytes": 6214181,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.0066122520001954,
+      "bytes": 1096551,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.3737143940015812,
-      "bytes": 10463664,
-      "pages": 36,
+      "seconds": 1.1023049610012094,
+      "bytes": 10542150,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.8019685460021719,
-      "bytes": 2186825,
-      "pages": 23,
-      "shards": 2
+      "seconds": 0.9890626239939593,
+      "bytes": 7730044,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.2952469239971833,
-      "bytes": 12034549,
-      "pages": 44,
-      "shards": 3
+      "seconds": 0.8456406199984485,
+      "bytes": 955055,
+      "pages": 28,
+      "shards": 1
     },
     {
-      "seconds": 0.8924442770003225,
-      "bytes": 6155396,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2024492920027114,
+      "bytes": 7626352,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.1965257189949625,
-      "bytes": 6621394,
-      "pages": 31,
-      "shards": 4
+      "seconds": 0.9993227560044033,
+      "bytes": 10754654,
+      "pages": 25,
+      "shards": 3
     },
     {
-      "seconds": 1.1455287000062526,
-      "bytes": 6484150,
+      "seconds": 0.9575367360012024,
+      "bytes": 6150393,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.8978742559993407,
-      "bytes": 2461928,
+      "seconds": 1.0901530560004176,
+      "bytes": 2315430,
       "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.9092525360028958,
-      "bytes": 6015320,
-      "pages": 26,
+      "seconds": 0.974858529996709,
+      "bytes": 6294959,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3533267960010562,
-      "bytes": 11703451,
-      "pages": 42,
+      "seconds": 1.211790827997902,
+      "bytes": 10339350,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.2227107450016774,
-      "bytes": 7523214,
-      "pages": 33,
+      "seconds": 1.1235615409968887,
+      "bytes": 2445045,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1649940130009782,
+      "bytes": 11872287,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0604035779979313,
-      "bytes": 6119324,
-      "pages": 36,
+      "seconds": 1.1487145990031422,
+      "bytes": 2742520,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.0826924209977733,
-      "bytes": 902486,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.1007023089987342,
+      "bytes": 9561100,
+      "pages": 32,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857
-size 721765055
+oid sha256:26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725
+size 721762815

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": 61,
+  "prefetch_low_water": 30,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,