Qwen90M step 415000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 410000,
+  "step": 415000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5",
+    "model.safetensors": "d68a25458859892a42235b6e969f63914c1b99b2fd338f2b890bb7fa60b747e4",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc",
-    "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
+    "trainer.pt": "acd7ce5336627e5ce5fa8de05c0c11732ac9fb116efd448705a06cecfb43e314",
+    "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
     "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
-    "shard_scheduler.json": "dcd8a9cb4cda262678652187e892a303cca733af30618ef17353af63ba53c942",
-    "coverage.json": "551c8b6a0d7c9ec77f8c7a79a16a6436443df8715ce046c0113917bd6f5dc3f4",
+    "shard_scheduler.json": "305f8ca3d6091cb9858cc65c58e7bb937f6a7682dcf2f326d88836f842db83af",
+    "coverage.json": "a7a33dedd6cfa71fd8d0e341d68024ea27360dfb5b66bdeb4084edbfa78d45b0",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 410000,
+  "step": 415000,
   "stats": {
-    "climbmix/blocks": 14760000,
-    "climbmix/targets": 24673977216,
-    "climbmix/input_tokens": 24675540036,
-    "ultra_qa/blocks": 5535000,
-    "ultra_qa/targets": 9029200380,
-    "ultra_qa/input_tokens": 9029217755,
-    "cortex/blocks": 5535000,
-    "cortex/targets": 210523889,
-    "cortex/input_tokens": 8798719668,
-    "ultra_style/blocks": 5535000,
-    "ultra_style/targets": 8755304288,
-    "ultra_style/input_tokens": 8755763253,
-    "openmath/blocks": 3690000,
-    "openmath/targets": 5702199677,
-    "openmath/input_tokens": 5704099937,
-    "rewrite6/blocks": 1845000,
-    "rewrite6/targets": 3024643728,
-    "rewrite6/input_tokens": 3027959756
+    "climbmix/blocks": 14940000,
+    "climbmix/targets": 24974931240,
+    "climbmix/input_tokens": 24976513176,
+    "ultra_qa/blocks": 5602500,
+    "ultra_qa/targets": 9139274938,
+    "ultra_qa/input_tokens": 9139292497,
+    "cortex/blocks": 5602500,
+    "cortex/targets": 213083506,
+    "cortex/input_tokens": 8905936709,
+    "ultra_style/blocks": 5602500,
+    "ultra_style/targets": 8862189935,
+    "ultra_style/input_tokens": 8862654526,
+    "openmath/blocks": 3735000,
+    "openmath/targets": 5771635114,
+    "openmath/input_tokens": 5773558303,
+    "rewrite6/blocks": 1867500,
+    "rewrite6/targets": 3061493752,
+    "rewrite6/input_tokens": 3064850276
   },
   "sources": {
     "climbmix": {
-      "page": 2021,
-      "row": 1375,
+      "page": 2046,
+      "row": 197,
       "start": {
-        "file": 97,
+        "file": 99,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 69,
+          "row_group": 3,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 8279391
+      "offset": 0,
+      "documents": 8380613
     },
     "rewrite6": {
-      "page": 50,
-      "row": 510,
+      "page": 14,
+      "row": 193,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 204800
+          "row": 57344
         }
       },
-      "epoch": 22,
+      "epoch": 23,
       "offset": 0,
-      "documents": 4724198
+      "documents": 4781829
     },
     "openmath": {
-      "page": 1263,
-      "row": 1659,
+      "page": 1278,
+      "row": 2960,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 370,
-          "row": 98
+          "row_group": 431,
+          "row": 538
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 5174907
+      "documents": 5237648
     },
     "ultra_qa": {
-      "page": 546,
-      "row": 940,
+      "page": 552,
+      "row": 3630,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 2,
-          "row": 28720
+          "row": 53296
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2237356
+      "documents": 2264622
     },
     "ultra_style": {
-      "page": 833,
-      "row": 2801,
+      "page": 843,
+      "row": 3466,
       "start": {
-        "file": 4,
+        "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 76397
+          "row_group": 0,
+          "row": 28216
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 3414769
+      "offset": 2048,
+      "documents": 3456394
     },
     "cortex": {
-      "page": 107353,
-      "row": 26,
+      "page": 108662,
+      "row": 13,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 13741210
+      "documents": 13908749
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 410000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "dcd8a9cb4cda262678652187e892a303cca733af30618ef17353af63ba53c942", "coverage.json": "551c8b6a0d7c9ec77f8c7a79a16a6436443df8715ce046c0113917bd6f5dc3f4", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000410000"}
+{"step": 415000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "d68a25458859892a42235b6e969f63914c1b99b2fd338f2b890bb7fa60b747e4", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "acd7ce5336627e5ce5fa8de05c0c11732ac9fb116efd448705a06cecfb43e314", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "305f8ca3d6091cb9858cc65c58e7bb937f6a7682dcf2f326d88836f842db83af", "coverage.json": "a7a33dedd6cfa71fd8d0e341d68024ea27360dfb5b66bdeb4084edbfa78d45b0", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000415000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5
+oid sha256:d68a25458859892a42235b6e969f63914c1b99b2fd338f2b890bb7fa60b747e4
 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": 76595.74770061958,
-      "required_pages_per_second_upper_estimate": 0.005299071552549478
+      "required_token_positions_per_second": 76122.78800998902,
+      "required_pages_per_second_upper_estimate": 0.005266351103733979
     },
     {
       "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": 9574.468462577448,
-      "required_pages_per_second_upper_estimate": 0.004008630007644012
+      "required_token_positions_per_second": 9515.348501248627,
+      "required_pages_per_second_upper_estimate": 0.003983877714400815
     },
     {
       "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": 19148.936925154896,
-      "required_pages_per_second_upper_estimate": 0.0039882185700838456
+      "required_token_positions_per_second": 19030.697002497254,
+      "required_pages_per_second_upper_estimate": 0.003963592312390709
     },
     {
       "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": 28723.405387732342,
-      "required_pages_per_second_upper_estimate": 0.0017219242948939932
+      "required_token_positions_per_second": 28546.045503745878,
+      "required_pages_per_second_upper_estimate": 0.001711291840661867
     },
     {
       "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": 28723.405387732342,
-      "required_pages_per_second_upper_estimate": 0.002233118427988008
+      "required_token_positions_per_second": 28546.045503745878,
+      "required_pages_per_second_upper_estimate": 0.0022193294771317438
     },
     {
       "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": 28723.405387732342,
-      "required_pages_per_second_upper_estimate": 0.3684330035239715
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28546.045503745878,
+      "required_pages_per_second_upper_estimate": 0.3661580213663996
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.09953904679677805,
-    "producer_p95_seconds": 0.22590240799945605,
-    "gpu_update_seconds": 0.9625599620513087,
-    "producer_batches_per_second": 10.046308782136828,
-    "gpu_batches_per_second": 1.0388963175539765,
-    "producer_headroom": 9.670174600089354,
+    "producer_batch_seconds": 0.06699381209364219,
+    "producer_p95_seconds": 0.2311606559997017,
+    "gpu_update_seconds": 0.9685404584803862,
+    "producer_batches_per_second": 14.926751721520583,
+    "gpu_batches_per_second": 1.0324813911945123,
+    "producer_headroom": 14.457162955984439,
     "sustainable": true,
-    "buffer_seconds": 60.641277609232446,
+    "buffer_seconds": 60.04950842578394,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 352991.2228158892,
-    "upload_bytes_per_second": 5469736.43334308,
+    "generation_bytes_per_second": 403752.13096224185,
+    "upload_bytes_per_second": 5602505.963495252,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 75.74369920117863,
+    "estimated_full_shards_per_hour": 86.6358084359211,
     "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": 1.0353907819999222,
-      "bytes": 2349590,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.168341065000277,
+      "bytes": 9926928,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.3265869809993092,
-      "bytes": 11233813,
-      "pages": 40,
-      "shards": 3
+      "seconds": 0.8547535799989419,
+      "bytes": 1160518,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.9781427679999979,
-      "bytes": 8201204,
-      "pages": 34,
+      "seconds": 0.9846142869992036,
+      "bytes": 7709639,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9253323299999465,
-      "bytes": 6468324,
-      "pages": 36,
+      "seconds": 1.0955427600001713,
+      "bytes": 6199869,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0003412309997657,
-      "bytes": 1020572,
-      "pages": 29,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2060352290000083,
-      "bytes": 11046742,
-      "pages": 40,
+      "seconds": 1.6065525880003406,
+      "bytes": 12580162,
+      "pages": 33,
       "shards": 4
     },
     {
-      "seconds": 1.0752750450001258,
-      "bytes": 6242403,
-      "pages": 32,
+      "seconds": 0.995714869000949,
+      "bytes": 5809499,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.8482705019996502,
-      "bytes": 2583323,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.7974183579990495,
+      "bytes": 1155225,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.3519573450003008,
-      "bytes": 11517949,
-      "pages": 22,
+      "seconds": 1.29576844300027,
+      "bytes": 7786718,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.325917474000562,
-      "bytes": 11191562,
-      "pages": 43,
+      "seconds": 1.0071225869996852,
+      "bytes": 672738,
+      "pages": 21,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3492830719987978,
+      "bytes": 12380695,
+      "pages": 48,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1991475859995262,
+      "bytes": 11620692,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.1523795839993909,
-      "bytes": 2573171,
+      "seconds": 0.9693775400010054,
+      "bytes": 6039517,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.718842794000011,
-      "bytes": 5798754,
-      "pages": 27,
+      "seconds": 0.8906193579987303,
+      "bytes": 2228503,
+      "pages": 22,
       "shards": 2
     },
     {
-      "seconds": 1.0535618110006908,
-      "bytes": 1088694,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.1276412609986437,
+      "bytes": 11020284,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.3156409709999934,
-      "bytes": 11593993,
+      "seconds": 0.8274413319995801,
+      "bytes": 1262070,
       "pages": 35,
-      "shards": 4
+      "shards": 1
     },
     {
-      "seconds": 2.0578567419997853,
-      "bytes": 11618530,
-      "pages": 30,
+      "seconds": 1.0886287620014627,
+      "bytes": 7651274,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2497664069996972,
-      "bytes": 1138477,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.1375053659994592,
+      "bytes": 11652818,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.3184527379999054,
-      "bytes": 7724848,
-      "pages": 34,
+      "seconds": 2.7600006199991185,
+      "bytes": 6232730,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9524561609996454,
-      "bytes": 5747185,
-      "pages": 38,
+      "seconds": 1.0662739279996458,
+      "bytes": 6304947,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0773643330003324,
-      "bytes": 7156662,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3278628770003706,
-      "bytes": 6286061,
+      "seconds": 1.1643560489992524,
+      "bytes": 6319425,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9090305769996121,
-      "bytes": 7087614,
-      "pages": 37,
+      "seconds": 2.530195315999663,
+      "bytes": 2235467,
+      "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 1.3239770070003942,
-      "bytes": 12583619,
+      "seconds": 1.0548529980005696,
+      "bytes": 10543433,
       "pages": 38,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 0.9289323200000581,
-      "bytes": 883918,
-      "pages": 25,
-      "shards": 2
+      "seconds": 1.3102050620000227,
+      "bytes": 8381851,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.464534527000069,
-      "bytes": 7593844,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.10793444599949,
+      "bytes": 6143663,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.0184807959994941,
-      "bytes": 301114,
+      "seconds": 0.9255875790004211,
+      "bytes": 259525,
       "pages": 8,
       "shards": 1
     },
     {
-      "seconds": 1.0933929989987519,
-      "bytes": 5793673,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.9178394819991809,
+      "bytes": 643385,
+      "pages": 18,
+      "shards": 1
     },
     {
-      "seconds": 1.0935950399998546,
-      "bytes": 4381259,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.5178640079993784,
+      "bytes": 12104268,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.3099932099994476,
-      "bytes": 13324819,
-      "pages": 42,
-      "shards": 4
+      "seconds": 0.9816815789999964,
+      "bytes": 6359822,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9060626449991105,
-      "bytes": 6034341,
-      "pages": 27,
+      "seconds": 1.1105975430000399,
+      "bytes": 2661454,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1673664749996533,
-      "bytes": 6863070,
-      "pages": 38,
+      "seconds": 1.082202919998963,
+      "bytes": 11730331,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 2.7079190219992597,
-      "bytes": 5909001,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.0211684169989894,
+      "bytes": 10395259,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1004524960007984,
-      "bytes": 6145895,
+      "seconds": 0.8835518189989671,
+      "bytes": 2669717,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9040275779989315,
-      "bytes": 2336556,
+      "seconds": 1.406538745000944,
+      "bytes": 5882576,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.34454566200111,
-      "bytes": 16421464,
-      "pages": 42,
-      "shards": 4
+      "seconds": 1.4825863509995543,
+      "bytes": 11209775,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 0.9113021489993116,
-      "bytes": 908789,
-      "pages": 27,
-      "shards": 1
+      "seconds": 0.9785202229995775,
+      "bytes": 2340420,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 0.963736620000418,
-      "bytes": 7684006,
-      "pages": 38,
+      "seconds": 1.288048930999139,
+      "bytes": 12112969,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1074004389993206,
-      "bytes": 6213956,
-      "pages": 36,
+      "seconds": 1.0779387669990683,
+      "bytes": 2658914,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0808812039995246,
-      "bytes": 6459806,
-      "pages": 30,
+      "seconds": 0.9456378320010117,
+      "bytes": 5728053,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.0713844300007622,
+      "bytes": 10175048,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1259296549997089,
-      "bytes": 5894764,
+      "seconds": 1.0568985930003691,
+      "bytes": 2271584,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.93207276099929,
-      "bytes": 6831941,
-      "pages": 31,
+      "seconds": 1.734822353000709,
+      "bytes": 6223398,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0320983010005875,
-      "bytes": 7553870,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.168341065000277,
-      "bytes": 9926928,
-      "pages": 30,
+      "seconds": 1.2489348130002327,
+      "bytes": 8291174,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 0.8547535799989419,
-      "bytes": 1160518,
-      "pages": 33,
-      "shards": 1
-    },
-    {
-      "seconds": 0.9846142869992036,
-      "bytes": 7709639,
-      "pages": 38,
+      "seconds": 1.122991837999507,
+      "bytes": 10369739,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0955427600001713,
-      "bytes": 6199869,
+      "seconds": 1.367829029000859,
+      "bytes": 6101351,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.6065525880003406,
-      "bytes": 12580162,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.9580737389987917,
+      "bytes": 2350217,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 0.995714869000949,
-      "bytes": 5809499,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.1762084339989087,
+      "bytes": 10342096,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 0.7974183579990495,
-      "bytes": 1155225,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.164775548999387,
+      "bytes": 5765223,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.29576844300027,
-      "bytes": 7786718,
-      "pages": 38,
+      "seconds": 0.9962533819998498,
+      "bytes": 8524201,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.0071225869996852,
-      "bytes": 672738,
-      "pages": 21,
-      "shards": 1
+      "seconds": 1.0446712169996317,
+      "bytes": 5880079,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.3492830719987978,
-      "bytes": 12380695,
-      "pages": 48,
+      "seconds": 1.5804328610010998,
+      "bytes": 12088466,
+      "pages": 42,
       "shards": 4
     },
     {
-      "seconds": 1.1991475859995262,
-      "bytes": 11620692,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.7921914839989768,
+      "bytes": 902002,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 0.9693775400010054,
-      "bytes": 6039517,
-      "pages": 37,
+      "seconds": 1.0630905629986955,
+      "bytes": 6092917,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.8906193579987303,
-      "bytes": 2228503,
-      "pages": 22,
+      "seconds": 0.9817484880004486,
+      "bytes": 2589304,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1276412609986437,
-      "bytes": 11020284,
+      "seconds": 1.2340881659984007,
+      "bytes": 16290946,
       "pages": 38,
       "shards": 4
     },
     {
-      "seconds": 0.8274413319995801,
-      "bytes": 1262070,
+      "seconds": 1.10198693500206,
+      "bytes": 7334244,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.7830869729987171,
+      "bytes": 1161196,
       "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.0886287620014627,
-      "bytes": 7651274,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.0245182449980348,
+      "bytes": 6075842,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1375053659994592,
-      "bytes": 11652818,
+      "seconds": 1.3879416819981998,
+      "bytes": 11260844,
       "pages": 36,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 2.7600006199991185,
-      "bytes": 6232730,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.0456256510005915,
+      "bytes": 862691,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.0662739279996458,
-      "bytes": 6304947,
-      "pages": 36,
-      "shards": 2
+      "seconds": 2.013755919000687,
+      "bytes": 11740969,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.1643560489992524,
-      "bytes": 6319425,
-      "pages": 36,
+      "seconds": 0.9664811299990106,
+      "bytes": 2719675,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 2.530195315999663,
-      "bytes": 2235467,
-      "pages": 24,
-      "shards": 2
+      "seconds": 1.2011403209980926,
+      "bytes": 11095830,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.0548529980005696,
-      "bytes": 10543433,
-      "pages": 38,
+      "seconds": 1.0495139129998279,
+      "bytes": 7453354,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.3102050620000227,
-      "bytes": 8381851,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.6562018110016652,
+      "bytes": 1121250,
+      "pages": 33,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc
-size 721789695
+oid sha256:acd7ce5336627e5ce5fa8de05c0c11732ac9fb116efd448705a06cecfb43e314
+size 721777151

training_config.json

@@ -46,7 +46,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
+  "prefetch_batches": 62,
   "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,