Qwen90M step 360000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 355000,
+  "step": 360000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "e07e1e6496a6b48cf11640c7168b06232c8e131cc65408e66ebbe90be5fcabf0",
+    "model.safetensors": "aefca4aceeb21f90da9c36c76bd5a05ccb0e95d3e2344dbae626f2f7f0fcc916",
     "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",
+    "trainer.pt": "ef5519c0d2203b921c794e8e2cec4418be2e30e6f0ee6e640ab5ca3fc6c954f5",
+    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
+    "shard_scheduler.json": "83f1c90fc05480fdbcecae8e2db98bf4bb3dc03eaa8272164f0d6fe631c29fab",
+    "coverage.json": "f2b2198c10a4d312de6292dca73b81effe885572010ec59de71866b8e4bd577d",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 355000,
+  "step": 360000,
   "stats": {
-    "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
+    "climbmix/blocks": 12960000,
+    "climbmix/targets": 21664115367,
+    "climbmix/input_tokens": 21665488277,
+    "ultra_qa/blocks": 4860000,
+    "ultra_qa/targets": 7928409470,
+    "ultra_qa/input_tokens": 7928424783,
+    "cortex/blocks": 4860000,
+    "cortex/targets": 184879230,
+    "cortex/input_tokens": 7725738038,
+    "ultra_style/blocks": 4860000,
+    "ultra_style/targets": 7687434969,
+    "ultra_style/input_tokens": 7687838419,
+    "openmath/blocks": 3240000,
+    "openmath/targets": 5006598820,
+    "openmath/input_tokens": 5008267424,
+    "rewrite6/blocks": 1620000,
+    "rewrite6/targets": 2655769484,
+    "rewrite6/input_tokens": 2658681060
   },
   "sources": {
     "climbmix": {
-      "page": 1750,
-      "row": 2459,
+      "page": 1775,
+      "row": 1037,
       "start": {
-        "file": 84,
+        "file": 85,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 58,
+          "row_group": 75,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 7170459
+      "documents": 7271437
     },
     "rewrite6": {
-      "page": 45,
-      "row": 3686,
+      "page": 9,
+      "row": 3028,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 184320
+          "row": 36864
         }
       },
-      "epoch": 19,
+      "epoch": 20,
       "offset": 0,
-      "documents": 4090682
+      "documents": 4147972
     },
     "openmath": {
-      "page": 1094,
-      "row": 149,
+      "page": 1109,
+      "row": 1368,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 114,
-          "row": 524
+          "row_group": 175,
+          "row": 964
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 4481173
+      "documents": 4543832
     },
     "ultra_qa": {
-      "page": 472,
-      "row": 3902,
+      "page": 479,
+      "row": 2483,
       "start": {
         "file": 3,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 51451
+          "row_group": 6,
+          "row": 15507
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1937214
+      "offset": 4096,
+      "documents": 1964467
     },
     "ultra_style": {
-      "page": 722,
-      "row": 802,
+      "page": 732,
+      "row": 901,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 2,
-          "row": 47149
+          "row_group": 3,
+          "row": 3439
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 2958114
+      "offset": 2048,
+      "documents": 2999173
     },
     "cortex": {
-      "page": 92951,
-      "row": 78,
+      "page": 94259,
+      "row": 79,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 11897806
+      "documents": 12065231
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"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"}
+{"step": 360000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "aefca4aceeb21f90da9c36c76bd5a05ccb0e95d3e2344dbae626f2f7f0fcc916", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ef5519c0d2203b921c794e8e2cec4418be2e30e6f0ee6e640ab5ca3fc6c954f5", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "83f1c90fc05480fdbcecae8e2db98bf4bb3dc03eaa8272164f0d6fe631c29fab", "coverage.json": "f2b2198c10a4d312de6292dca73b81effe885572010ec59de71866b8e4bd577d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000360000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e07e1e6496a6b48cf11640c7168b06232c8e131cc65408e66ebbe90be5fcabf0
+oid sha256:aefca4aceeb21f90da9c36c76bd5a05ccb0e95d3e2344dbae626f2f7f0fcc916
 size 360622744

performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 86,
     "mode": "compiled",
-    "microbatch_seconds": 0.9004576890000067,
+    "microbatch_seconds": 0.9009280120000085,
     "peak_gb": 85.39913082122803,
     "fits_budget": true,
-    "estimated_update_seconds": 0.9568634569999972,
-    "input_tokens_per_second": 195598.30756245417
+    "estimated_update_seconds": 0.9565700770000376,
+    "input_tokens_per_second": 195496.19687038695
   },
   "measurements": [
     {

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": 73436.5976740253,
-      "required_pages_per_second_upper_estimate": 0.006365298163029331
+      "required_token_positions_per_second": 75665.69836686787,
+      "required_pages_per_second_upper_estimate": 0.00655851096148087
     },
     {
       "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": 9179.574709253162,
-      "required_pages_per_second_upper_estimate": 0.003661869598393634
+      "required_token_positions_per_second": 9458.212295858484,
+      "required_pages_per_second_upper_estimate": 0.003773022297693667
     },
     {
       "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": 18359.149418506324,
-      "required_pages_per_second_upper_estimate": 0.0042470661079772485
+      "required_token_positions_per_second": 18916.42459171697,
+      "required_pages_per_second_upper_estimate": 0.004375981911591458
     },
     {
       "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": 27538.724127759484,
-      "required_pages_per_second_upper_estimate": 0.0016769638148614063
+      "required_token_positions_per_second": 28374.636887575452,
+      "required_pages_per_second_upper_estimate": 0.0017278665162388923
     },
     {
       "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": 27538.724127759484,
-      "required_pages_per_second_upper_estimate": 0.00212000721542238
+      "required_token_positions_per_second": 28374.636887575452,
+      "required_pages_per_second_upper_estimate": 0.002184358093627632
     },
     {
       "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": 21,
-      "required_token_positions_per_second": 27538.724127759484,
-      "required_pages_per_second_upper_estimate": 0.3419387874859938
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28374.636887575452,
+      "required_pages_per_second_upper_estimate": 0.3523180263428667
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,149 @@
   "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": 60,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
-    "batch_memory_estimate_bytes": 690670,
-    "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,
+    "batch_memory_estimate_bytes": 717798,
+    "producer_batch_seconds": 0.13129274778125932,
+    "producer_p95_seconds": 0.1682472989996313,
+    "gpu_update_seconds": 0.9743913238271736,
+    "producer_batches_per_second": 7.61656692314836,
+    "gpu_batches_per_second": 1.026281716130478,
+    "producer_headroom": 7.421516727264794,
     "sustainable": true,
-    "buffer_seconds": 60.23808482571717,
+    "buffer_seconds": 60.412262077284765,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 324828.29635209485,
-    "upload_bytes_per_second": 5287585.875586291,
+    "generation_bytes_per_second": 351443.47014520573,
+    "upload_bytes_per_second": 5528607.3444076935,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 69.70059078142295,
+    "estimated_full_shards_per_hour": 75.41158750788811,
     "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": 0.9463060489979398,
-      "bytes": 743440,
-      "pages": 25,
+      "seconds": 0.8837165979998645,
+      "bytes": 52118,
+      "pages": 1,
       "shards": 1
     },
     {
-      "seconds": 1.0638403249977273,
-      "bytes": 11964464,
-      "pages": 46,
-      "shards": 4
-    },
-    {
-      "seconds": 1.8820245999959297,
-      "bytes": 865291,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3777121140010422,
-      "bytes": 7459243,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 2.28389226100262,
-      "bytes": 15733663,
-      "pages": 34,
-      "shards": 5
-    },
-    {
-      "seconds": 0.7980627049983013,
-      "bytes": 1094775,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0217303430035827,
-      "bytes": 7986291,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8759991519982577,
-      "bytes": 805335,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0561498970055254,
-      "bytes": 10650758,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2883731410038308,
-      "bytes": 7659801,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1231011929994565,
-      "bytes": 6897324,
+      "seconds": 1.4180460719999246,
+      "bytes": 7624068,
       "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0080412499955855,
-      "bytes": 7466769,
-      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.1737449919965002,
-      "bytes": 4963940,
-      "pages": 26,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0047582629995304,
-      "bytes": 6204635,
-      "pages": 42,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2032458940011566,
-      "bytes": 7462591,
+      "seconds": 1.001895477999824,
+      "bytes": 1118907,
       "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0979009180009598,
-      "bytes": 1059989,
-      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.474424235006154,
-      "bytes": 14927789,
-      "pages": 21,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0885418239995488,
-      "bytes": 8199005,
-      "pages": 47,
+      "seconds": 1.186168517999704,
+      "bytes": 7790528,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9340340310009196,
-      "bytes": 958542,
+      "seconds": 1.008963309000137,
+      "bytes": 10351006,
       "pages": 30,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0266037770052208,
-      "bytes": 7314912,
-      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1214634670031955,
-      "bytes": 5616787,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0272974420004175,
-      "bytes": 6127988,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1608434069930809,
-      "bytes": 13413721,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8657637560027069,
-      "bytes": 989817,
+      "seconds": 0.8917956289997164,
+      "bytes": 6042604,
       "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 0.8787106930030859,
-      "bytes": 615085,
-      "pages": 17,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1383813439970254,
-      "bytes": 11307353,
-      "pages": 35,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9193243289992097,
-      "bytes": 5971902,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "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": 0.8686913920028019,
-      "bytes": 2634908,
-      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1666307280029287,
-      "bytes": 5700254,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8491491380045773,
-      "bytes": 1087747,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1218249969970202,
-      "bytes": 11930213,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 1.2040022809960647,
-      "bytes": 6093346,
+      "seconds": 1.600639215999763,
+      "bytes": 2577832,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9920474090031348,
-      "bytes": 6801926,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4719860830009566,
-      "bytes": 7846458,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9881938749967958,
-      "bytes": 4661656,
-      "pages": 32,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2476324340022984,
-      "bytes": 7151046,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9890799559943844,
-      "bytes": 6390118,
+      "seconds": 0.8447169340001892,
+      "bytes": 6262278,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1548123519969522,
-      "bytes": 1183539,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2286420309974346,
-      "bytes": 17809213,
+      "seconds": 1.2238314549999814,
+      "bytes": 9984108,
       "pages": 32,
-      "shards": 5
+      "shards": 3
     },
     {
-      "seconds": 0.9948159990017302,
-      "bytes": 1198699,
+      "seconds": 0.8533821090004494,
+      "bytes": 2621863,
       "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0069344939984148,
-      "bytes": 6086180,
-      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.1158472850002,
-      "bytes": 7287693,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.026877347001573,
-      "bytes": 5561907,
-      "pages": 34,
+      "seconds": 1.0488426049996633,
+      "bytes": 5758950,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.067893715997343,
-      "bytes": 7222726,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3662753429962322,
-      "bytes": 11997986,
-      "pages": 39,
+      "seconds": 1.0318703240000104,
+      "bytes": 8301935,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 0.77281211099762,
-      "bytes": 859716,
-      "pages": 26,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1721114309984841,
-      "bytes": 11272182,
-      "pages": 38,
+      "seconds": 1.1579725150004379,
+      "bytes": 10569380,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 0.8626750620023813,
-      "bytes": 1117608,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 0.9938163240003632,
-      "bytes": 5819862,
-      "pages": 27,
+      "seconds": 1.112220934999641,
+      "bytes": 6262678,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.1086062760005007,
-      "bytes": 7618244,
+      "seconds": 0.9345845090001603,
+      "bytes": 2621511,
       "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1348141839989694,
-      "bytes": 11320691,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 0.981941227997595,
-      "bytes": 7228716,
-      "pages": 30,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 0.914243400999112,
-      "bytes": 1165448,
-      "pages": 36,
-      "shards": 1
+      "seconds": 0.8686625299997104,
+      "bytes": 5945555,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.0446478919984656,
-      "bytes": 5999613,
-      "pages": 36,
+      "seconds": 1.1098429370003942,
+      "bytes": 4826131,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.4630355209956178,
-      "bytes": 11583159,
-      "pages": 30,
+      "seconds": 1.7090945019999708,
+      "bytes": 13118107,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 1.1508738010015804,
-      "bytes": 1166291,
+      "seconds": 1.577916964999531,
+      "bytes": 6095231,
       "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4934369809998316,
-      "bytes": 13370474,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8853696480000508,
-      "bytes": 1220951,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 2.1309150849992875,
-      "bytes": 9524825,
-      "pages": 30,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.1543053290006355,
-      "bytes": 7653463,
-      "pages": 39,
+      "seconds": 1.0922540970004775,
+      "bytes": 6780784,
+      "pages": 29,
       "shards": 3
-    },
-    {
-      "seconds": 1.0381377000012435,
-      "bytes": 839375,
-      "pages": 26,
-      "shards": 1
     }
   ],
   "cache_key": {
@@ -606,7 +342,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    2.358020696999972,
-    0.013266629999975521
+    3.301654667999969,
+    0.11550889399995867
   ]
 }

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27deb3f82952932a7dca09467002556e1f8621d42020878ad6f833309e2c4d60
-size 721770687
+oid sha256:ef5519c0d2203b921c794e8e2cec4418be2e30e6f0ee6e640ab5ca3fc6c954f5
+size 721774271

training_config.json

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