Qwen90M step 270000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 265000,
+  "step": 270000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "ee0020789a0e214a0997c223fbd46efde8cdecccc05fb984287e0e416e538818",
+    "model.safetensors": "ff80e0adb68dde4f1a64a84ef34645bbab4563c2ba6a6e22e8ecc9e7f0d64151",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "351fececeaaf7d61cfa47cb55a4c1fc10dd091aef0f1465531852527c3727ddd",
-    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "trainer.pt": "daa4557c6d9877c2d582a143485f465f77ac3654b5943b52633de0782589f2a9",
+    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
     "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
-    "shard_scheduler.json": "2f2de5af2999d7b5fe09209f6bbf509804ef8b94f6516c38e9e688d203d65511",
-    "coverage.json": "c14d992a21f757f9e0584bc908a2da7fa8b6a7ebe5a0d5d6d02f098aff871438",
+    "shard_scheduler.json": "af7ddb35dfaf68caf6ea156cec645638aef2aea57e28dc84fbde94349cb2bc7f",
+    "coverage.json": "62a499fb026b3c3a39bc48e383984377f6842cc6452ea627930a87e4ce594788",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 265000,
+  "step": 270000,
   "stats": {
-    "climbmix/blocks": 9540000,
-    "climbmix/targets": 15947557328,
-    "climbmix/input_tokens": 15948568353,
-    "ultra_qa/blocks": 3577500,
-    "ultra_qa/targets": 5836258589,
-    "ultra_qa/input_tokens": 5836269832,
-    "cortex/blocks": 3577500,
-    "cortex/targets": 136096913,
-    "cortex/input_tokens": 5686718997,
-    "ultra_style/blocks": 3577500,
-    "ultra_style/targets": 5659331784,
-    "ultra_style/input_tokens": 5659628550,
-    "openmath/blocks": 2385000,
-    "openmath/targets": 3685319164,
-    "openmath/input_tokens": 3686547926,
-    "rewrite6/blocks": 1192500,
-    "rewrite6/targets": 1954941231,
-    "rewrite6/input_tokens": 1957084611
+    "climbmix/blocks": 9720000,
+    "climbmix/targets": 16248306295,
+    "climbmix/input_tokens": 16249336141,
+    "ultra_qa/blocks": 3645000,
+    "ultra_qa/targets": 5946191498,
+    "ultra_qa/input_tokens": 5946202929,
+    "cortex/blocks": 3645000,
+    "cortex/targets": 138670584,
+    "cortex/input_tokens": 5793992704,
+    "ultra_style/blocks": 3645000,
+    "ultra_style/targets": 5765190778,
+    "ultra_style/input_tokens": 5765493512,
+    "openmath/blocks": 2430000,
+    "openmath/targets": 3754982954,
+    "openmath/input_tokens": 3756234898,
+    "rewrite6/blocks": 1215000,
+    "rewrite6/targets": 1991834205,
+    "rewrite6/input_tokens": 1994017932
   },
   "sources": {
     "climbmix": {
-      "page": 1306,
-      "row": 2467,
+      "page": 1331,
+      "row": 746,
       "start": {
-        "file": 63,
+        "file": 64,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 18,
+          "row_group": 35,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 5351843
+      "offset": 12288,
+      "documents": 5452522
     },
     "rewrite6": {
-      "page": 43,
-      "row": 1780,
+      "page": 7,
+      "row": 1260,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 176128
+          "row": 28672
         }
       },
-      "epoch": 14,
-      "offset": 2048,
-      "documents": 3053564
+      "epoch": 15,
+      "offset": 0,
+      "documents": 3110992
     },
     "openmath": {
-      "page": 816,
-      "row": 3499,
+      "page": 832,
+      "row": 915,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 285,
-          "row": 786
+          "row_group": 351,
+          "row": 322
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 3345835
+      "documents": 3408787
     },
     "ultra_qa": {
-      "page": 353,
-      "row": 292,
+      "page": 359,
+      "row": 2968,
       "start": {
         "file": 2,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 19137
+          "row": 43713
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 1446180
+      "documents": 1473432
     },
     "ultra_style": {
-      "page": 538,
-      "row": 3285,
+      "page": 549,
+      "row": 1375,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 63801
+          "row_group": 2,
+          "row": 23029
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2206933
+      "documents": 2250079
     },
     "cortex": {
-      "page": 69382,
-      "row": 24,
+      "page": 70690,
+      "row": 82,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 8880920
+      "documents": 9048402
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 265000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ee0020789a0e214a0997c223fbd46efde8cdecccc05fb984287e0e416e538818", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "351fececeaaf7d61cfa47cb55a4c1fc10dd091aef0f1465531852527c3727ddd", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "2f2de5af2999d7b5fe09209f6bbf509804ef8b94f6516c38e9e688d203d65511", "coverage.json": "c14d992a21f757f9e0584bc908a2da7fa8b6a7ebe5a0d5d6d02f098aff871438", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000265000"}
+{"step": 270000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ff80e0adb68dde4f1a64a84ef34645bbab4563c2ba6a6e22e8ecc9e7f0d64151", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "daa4557c6d9877c2d582a143485f465f77ac3654b5943b52633de0782589f2a9", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "af7ddb35dfaf68caf6ea156cec645638aef2aea57e28dc84fbde94349cb2bc7f", "coverage.json": "62a499fb026b3c3a39bc48e383984377f6842cc6452ea627930a87e4ce594788", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000270000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ee0020789a0e214a0997c223fbd46efde8cdecccc05fb984287e0e416e538818
+oid sha256:ff80e0adb68dde4f1a64a84ef34645bbab4563c2ba6a6e22e8ecc9e7f0d64151
 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": 75598.5316342505,
-      "required_pages_per_second_upper_estimate": 0.0065526891193301235
+      "required_token_positions_per_second": 74602.20012608299,
+      "required_pages_per_second_upper_estimate": 0.006466329629381285
     },
     {
       "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": 9449.816454281312,
-      "required_pages_per_second_upper_estimate": 0.003769673070959515
+      "required_token_positions_per_second": 9325.275015760373,
+      "required_pages_per_second_upper_estimate": 0.0037199916290730707
     },
     {
       "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": 18899.632908562624,
-      "required_pages_per_second_upper_estimate": 0.00437209745121723
+      "required_token_positions_per_second": 18650.550031520746,
+      "required_pages_per_second_upper_estimate": 0.004314476511322505
     },
     {
       "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": 28349.449362843934,
-      "required_pages_per_second_upper_estimate": 0.0017263327281314721
+      "required_token_positions_per_second": 27975.825047281116,
+      "required_pages_per_second_upper_estimate": 0.0017035809675689838
     },
     {
       "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": 28349.449362843934,
-      "required_pages_per_second_upper_estimate": 0.002182419088250269
+      "required_token_positions_per_second": 27975.825047281116,
+      "required_pages_per_second_upper_estimate": 0.0021536564541799422
     },
     {
       "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": 28349.449362843934,
-      "required_pages_per_second_upper_estimate": 0.3520052815829238
+      "estimated_pages_for_buffer": 21,
+      "required_token_positions_per_second": 27975.825047281116,
+      "required_pages_per_second_upper_estimate": 0.3473661180237793
     }
   ],
   "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": 690670,
-    "producer_batch_seconds": 0.263509006851649,
-    "producer_p95_seconds": 0.32680167599755805,
-    "gpu_update_seconds": 0.9752570374871801,
-    "producer_batches_per_second": 3.794936696653343,
-    "gpu_batches_per_second": 1.0253707090148994,
-    "producer_headroom": 3.7010387202295245,
+    "producer_batch_seconds": 0.2140537983123636,
+    "producer_p95_seconds": 0.3658770080000977,
+    "gpu_update_seconds": 0.9882818452457767,
+    "producer_batches_per_second": 4.671722753271231,
+    "gpu_batches_per_second": 1.0118570980642765,
+    "producer_headroom": 4.616978783079572,
     "sustainable": true,
-    "buffer_seconds": 60.46593632420517,
+    "buffer_seconds": 60.285192559992375,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 372935.845468488,
-    "upload_bytes_per_second": 4393328.842840834,
+    "generation_bytes_per_second": 268198.67309396865,
+    "upload_bytes_per_second": 4767939.973542438,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 80.02335093537312,
+    "estimated_full_shards_per_hour": 57.54919190038962,
     "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": 1.0260517099995923,
-      "bytes": 2406820,
-      "pages": 27,
-      "shards": 2
+      "seconds": 2.1326607650007645,
+      "bytes": 1150711,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.0375558390005608,
-      "bytes": 11763695,
-      "pages": 40,
+      "seconds": 2.922968579998269,
+      "bytes": 7646925,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.2267565980000654,
-      "bytes": 917900,
-      "pages": 27,
+      "seconds": 2.558325848000095,
+      "bytes": 890978,
+      "pages": 28,
       "shards": 1
     },
     {
-      "seconds": 1.2721305819977715,
-      "bytes": 12594000,
-      "pages": 39,
+      "seconds": 2.274885417002224,
+      "bytes": 15797221,
+      "pages": 34,
       "shards": 4
     },
     {
-      "seconds": 1.3089486960016075,
-      "bytes": 6052165,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.166924328001187,
-      "bytes": 2378485,
-      "pages": 30,
+      "seconds": 3.0481821479988866,
+      "bytes": 2407311,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.0251769999995304,
-      "bytes": 6275188,
+      "seconds": 1.2172690299994429,
+      "bytes": 6177200,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9176649480032211,
-      "bytes": 4600807,
-      "pages": 27,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5337045840024075,
-      "bytes": 13355543,
-      "pages": 42,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0210915280003974,
-      "bytes": 6074322,
+      "seconds": 1.8479619009995076,
+      "bytes": 5907251,
       "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.9218759530012903,
-      "bytes": 1088353,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2536445320001803,
-      "bytes": 11359516,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.345177949002391,
+      "bytes": 6941210,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.0509260079998057,
-      "bytes": 6234779,
+      "seconds": 1.2569710710013169,
+      "bytes": 6280484,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.154916157000116,
-      "bytes": 2367312,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.8538454429981357,
-      "bytes": 11918988,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.521943290001218,
+      "bytes": 13325077,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.6541002780031704,
-      "bytes": 6139021,
-      "pages": 33,
+      "seconds": 1.0711962380009936,
+      "bytes": 889212,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.9353761299971666,
-      "bytes": 7270095,
-      "pages": 29,
+      "seconds": 1.3221969680016628,
+      "bytes": 10781457,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 2.698638427998958,
-      "bytes": 6167283,
+      "seconds": 1.231661190999148,
+      "bytes": 2563678,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 2.6425060760011547,
-      "bytes": 2623268,
-      "pages": 37,
+      "seconds": 1.1248420010015252,
+      "bytes": 5951433,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 2.5355882220028434,
-      "bytes": 9765836,
-      "pages": 33,
+      "seconds": 1.6543720410008973,
+      "bytes": 7466727,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.361022408997087,
-      "bytes": 6242966,
-      "pages": 18,
+      "seconds": 1.3103912280021177,
+      "bytes": 10646431,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.2147977610002272,
-      "bytes": 6193785,
-      "pages": 35,
+      "seconds": 1.3748210780031513,
+      "bytes": 6494530,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9691767799995432,
-      "bytes": 2430413,
-      "pages": 30,
+      "seconds": 0.9434360529994592,
+      "bytes": 2587569,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0292037019971758,
-      "bytes": 5575102,
-      "pages": 18,
+      "seconds": 1.083952794000652,
+      "bytes": 5627431,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.288112906000606,
-      "bytes": 5599143,
-      "pages": 33,
+      "seconds": 1.2045630370012077,
+      "bytes": 11050050,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3764796949981246,
+      "bytes": 2478260,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0466477340014535,
-      "bytes": 7416772,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.1128817630014964,
+      "bytes": 6162674,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.263268196998979,
-      "bytes": 1135440,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2386194030004845,
+      "bytes": 6597417,
+      "pages": 26,
+      "shards": 2
     },
     {
-      "seconds": 1.37565966299735,
-      "bytes": 13270818,
+      "seconds": 1.1337901490005606,
+      "bytes": 11234450,
       "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 0.7815917669977352,
-      "bytes": 913950,
-      "pages": 28,
+      "seconds": 0.9795328820000577,
+      "bytes": 930296,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.2434416440009954,
-      "bytes": 10283141,
-      "pages": 41,
+      "seconds": 1.2613436630017532,
+      "bytes": 7588531,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.034428533999744,
-      "bytes": 7372684,
-      "pages": 29,
-      "shards": 3
+      "seconds": 0.8843855299965071,
+      "bytes": 5949979,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.070391503002611,
-      "bytes": 1123754,
+      "seconds": 1.1865619210002478,
+      "bytes": 5319050,
       "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.18286236700078,
-      "bytes": 5852988,
-      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.7732943640003214,
-      "bytes": 17450620,
-      "pages": 33,
-      "shards": 5
+      "seconds": 1.3752340050014027,
+      "bytes": 13115816,
+      "pages": 30,
+      "shards": 4
     },
     {
-      "seconds": 0.8563921040004061,
-      "bytes": 1024816,
+      "seconds": 1.269822567999654,
+      "bytes": 1251620,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.4308035579997522,
-      "bytes": 7307496,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.9414359830007015,
+      "bytes": 5894779,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.6164238569981535,
-      "bytes": 892499,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.1529198649986938,
+      "bytes": 11606253,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.2368712060015241,
-      "bytes": 10018327,
-      "pages": 37,
-      "shards": 3
+      "seconds": 0.9286228500022844,
+      "bytes": 849903,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.856085086998064,
-      "bytes": 7579977,
+      "seconds": 1.0139027430013812,
+      "bytes": 7512493,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2339968710002722,
-      "bytes": 6738522,
-      "pages": 29,
+      "seconds": 0.9002696629977436,
+      "bytes": 6383764,
+      "pages": 19,
       "shards": 2
     },
     {
-      "seconds": 2.4403519309998956,
-      "bytes": 5986840,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.1763414359993476,
+      "bytes": 9929012,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 2.588106572999095,
-      "bytes": 12282184,
+      "seconds": 1.3798298289984814,
+      "bytes": 7738436,
       "pages": 39,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 2.1326607650007645,
-      "bytes": 1150711,
-      "pages": 35,
+      "seconds": 0.956485711998539,
+      "bytes": 1216085,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 2.922968579998269,
-      "bytes": 7646925,
-      "pages": 36,
+      "seconds": 0.9746390660002362,
+      "bytes": 5704060,
+      "pages": 20,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1505642880001687,
+      "bytes": 6924827,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 2.558325848000095,
-      "bytes": 890978,
-      "pages": 28,
-      "shards": 1
+      "seconds": 0.9783257499984757,
+      "bytes": 5971368,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 2.274885417002224,
-      "bytes": 15797221,
-      "pages": 34,
+      "seconds": 1.5463005599995086,
+      "bytes": 13315083,
+      "pages": 40,
       "shards": 4
     },
     {
-      "seconds": 3.0481821479988866,
-      "bytes": 2407311,
-      "pages": 29,
+      "seconds": 1.3599293949991988,
+      "bytes": 4987866,
+      "pages": 33,
+      "shards": 3
+    },
+    {
+      "seconds": 1.145182983000268,
+      "bytes": 6203180,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.2172690299994429,
-      "bytes": 6177200,
+      "seconds": 0.9317472719994839,
+      "bytes": 2642711,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.8479619009995076,
-      "bytes": 5907251,
-      "pages": 30,
+      "seconds": 0.9964947309999843,
+      "bytes": 5697842,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.345177949002391,
-      "bytes": 6941210,
+      "seconds": 1.1368190850007522,
+      "bytes": 10318857,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.2569710710013169,
-      "bytes": 6280484,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.0222451929985255,
+      "bytes": 8375215,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.521943290001218,
-      "bytes": 13325077,
+      "seconds": 0.9998172479972709,
+      "bytes": 6024653,
       "pages": 36,
-      "shards": 4
+      "shards": 2
     },
     {
-      "seconds": 1.0711962380009936,
-      "bytes": 889212,
-      "pages": 26,
+      "seconds": 0.9568293169977551,
+      "bytes": 2417450,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3221969680016628,
-      "bytes": 10781457,
-      "pages": 38,
+      "seconds": 1.2115873000002466,
+      "bytes": 10372111,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.231661190999148,
-      "bytes": 2563678,
+      "seconds": 1.1448558080010116,
+      "bytes": 6260573,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1248420010015252,
-      "bytes": 5951433,
-      "pages": 29,
+      "seconds": 1.1358172180007386,
+      "bytes": 2620953,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.6543720410008973,
-      "bytes": 7466727,
-      "pages": 38,
+      "seconds": 1.4711359560023993,
+      "bytes": 11859463,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.3103912280021177,
-      "bytes": 10646431,
-      "pages": 28,
+      "seconds": 1.2465269989988883,
+      "bytes": 5535585,
+      "pages": 33,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2969893279987446,
+      "bytes": 7578437,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.3748210780031513,
-      "bytes": 6494530,
-      "pages": 37,
+      "seconds": 1.5002685959989321,
+      "bytes": 6193769,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 0.9434360529994592,
-      "bytes": 2587569,
-      "pages": 36,
+      "seconds": 1.1753300150012365,
+      "bytes": 2646901,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.083952794000652,
-      "bytes": 5627431,
+      "seconds": 1.4061265680029464,
+      "bytes": 9691299,
       "pages": 28,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0368577580011333,
+      "bytes": 6834427,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.2045630370012077,
-      "bytes": 11050050,
-      "pages": 39,
+      "seconds": 1.4560243180021644,
+      "bytes": 7277270,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.3764796949981246,
-      "bytes": 2478260,
-      "pages": 28,
+      "seconds": 1.4158115079990239,
+      "bytes": 6180346,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.1128817630014964,
-      "bytes": 6162674,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.5717087800003355,
+      "bytes": 560796,
+      "pages": 18,
+      "shards": 1
+    },
+    {
+      "seconds": 4.375384323997423,
+      "bytes": 11988381,
+      "pages": 43,
+      "shards": 4
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:351fececeaaf7d61cfa47cb55a4c1fc10dd091aef0f1465531852527c3727ddd
-size 721773055
+oid sha256:daa4557c6d9877c2d582a143485f465f77ac3654b5943b52633de0782589f2a9
+size 721768959

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,