Qwen90M step 595000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 590000,
+  "step": 595000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "2d5cd0fbe3e3ae4f47501307fc03d8c9a67b97fce7c8cd1d26c5286378e25c0e",
+    "model.safetensors": "947d55251820f157a3efb0410a2db9cfd0393475c09cc94095d69f9cbc248585",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "e6a73d2db07c024ebb237c7d0b6fcb773796b3b6ab1195a52d429d55bc86e73a",
-    "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb",
+    "trainer.pt": "04c2d546da542c12c58a0435f1ba889816d6f1028daf377c51a35cccae028102",
+    "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
     "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
-    "shard_scheduler.json": "44d308ae26c6e914df1a735812c742d18304f11ef2e74faf2202b5f592eda180",
-    "coverage.json": "33583797aa946780892ff16dcb005f14f82af20df711e2f0964d6f600512e644",
+    "shard_scheduler.json": "a238ae925bbc8d31568673428b064582cd86c23da8f0df6bda1ed49585c86033",
+    "coverage.json": "80387086e945d13c57f5522f96608f203c1992dacef7f48f7cc07e389650ef8a",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 590000,
+  "step": 595000,
   "stats": {
-    "climbmix/blocks": 21240000,
-    "climbmix/targets": 35505139268,
-    "climbmix/input_tokens": 35507389756,
-    "ultra_qa/blocks": 7965000,
-    "ultra_qa/targets": 12994238467,
-    "ultra_qa/input_tokens": 12994263542,
-    "cortex/blocks": 7965000,
-    "cortex/targets": 302869277,
-    "cortex/input_tokens": 12660969641,
-    "ultra_style/blocks": 7965000,
-    "ultra_style/targets": 12598185008,
-    "ultra_style/input_tokens": 12598846165,
-    "openmath/blocks": 5310000,
-    "openmath/targets": 8204944979,
-    "openmath/input_tokens": 8207677843,
-    "rewrite6/blocks": 2655000,
-    "rewrite6/targets": 4352536815,
-    "rewrite6/input_tokens": 4357308639
+    "climbmix/blocks": 21420000,
+    "climbmix/targets": 35806090870,
+    "climbmix/input_tokens": 35808360523,
+    "ultra_qa/blocks": 8032500,
+    "ultra_qa/targets": 13104381210,
+    "ultra_qa/input_tokens": 13104406506,
+    "cortex/blocks": 8032500,
+    "cortex/targets": 305435983,
+    "cortex/input_tokens": 12768242199,
+    "ultra_style/blocks": 8032500,
+    "ultra_style/targets": 12704953076,
+    "ultra_style/input_tokens": 12705619593,
+    "openmath/blocks": 5355000,
+    "openmath/targets": 8274625787,
+    "openmath/input_tokens": 8277382013,
+    "rewrite6/blocks": 2677500,
+    "rewrite6/targets": 4389363744,
+    "rewrite6/input_tokens": 4394176236
   },
   "sources": {
     "climbmix": {
-      "page": 2909,
-      "row": 1352,
+      "page": 2934,
+      "row": 289,
       "start": {
-        "file": 140,
+        "file": 141,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 64,
+          "row_group": 81,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11916616
+      "documents": 12017953
     },
     "rewrite6": {
-      "page": 4,
-      "row": 3444,
+      "page": 18,
+      "row": 3998,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 16384
+          "row": 73728
         }
       },
       "epoch": 33,
-      "offset": 2048,
-      "documents": 6798160
+      "offset": 0,
+      "documents": 6856058
     },
     "openmath": {
-      "page": 1817,
-      "row": 2899,
+      "page": 1833,
+      "row": 642,
       "start": {
         "file": 17,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 19,
-          "row": 382
+          "row_group": 84,
+          "row": 918
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 7445331
+      "offset": 2048,
+      "documents": 7508610
     },
     "ultra_qa": {
-      "page": 786,
-      "row": 115,
+      "page": 792,
+      "row": 2870,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 37192
+          "row": 61768
         }
       },
       "epoch": 0,
       "offset": 4096,
-      "documents": 3219571
+      "documents": 3246902
     },
     "ultra_style": {
-      "page": 1199,
-      "row": 3499,
+      "page": 1209,
+      "row": 3875,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 1,
-          "row": 31050
+          "row": 72010
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4914603
+      "documents": 4955939
     },
     "cortex": {
-      "page": 154475,
-      "row": 120,
+      "page": 155783,
+      "row": 98,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 19772920
+      "documents": 19940322
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 590000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2d5cd0fbe3e3ae4f47501307fc03d8c9a67b97fce7c8cd1d26c5286378e25c0e", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "e6a73d2db07c024ebb237c7d0b6fcb773796b3b6ab1195a52d429d55bc86e73a", "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "44d308ae26c6e914df1a735812c742d18304f11ef2e74faf2202b5f592eda180", "coverage.json": "33583797aa946780892ff16dcb005f14f82af20df711e2f0964d6f600512e644", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000590000"}
+{"step": 595000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "947d55251820f157a3efb0410a2db9cfd0393475c09cc94095d69f9cbc248585", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "04c2d546da542c12c58a0435f1ba889816d6f1028daf377c51a35cccae028102", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "a238ae925bbc8d31568673428b064582cd86c23da8f0df6bda1ed49585c86033", "coverage.json": "80387086e945d13c57f5522f96608f203c1992dacef7f48f7cc07e389650ef8a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000595000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2d5cd0fbe3e3ae4f47501307fc03d8c9a67b97fce7c8cd1d26c5286378e25c0e
+oid sha256:947d55251820f157a3efb0410a2db9cfd0393475c09cc94095d69f9cbc248585
 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": 77450.23149600206,
-      "required_pages_per_second_upper_estimate": 0.005358186724189602
+      "required_token_positions_per_second": 75999.11957643424,
+      "required_pages_per_second_upper_estimate": 0.005257795434550359
     },
     {
       "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": 9681.278937000257,
-      "required_pages_per_second_upper_estimate": 0.004053349322828503
+      "required_token_positions_per_second": 9499.88994705428,
+      "required_pages_per_second_upper_estimate": 0.003977405540570961
     },
     {
       "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": 19362.557874000515,
-      "required_pages_per_second_upper_estimate": 0.004032710180165127
+      "required_token_positions_per_second": 18999.77989410856,
+      "required_pages_per_second_upper_estimate": 0.003957153094052322
     },
     {
       "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": 29043.83681100077,
-      "required_pages_per_second_upper_estimate": 0.0017411336694484819
+      "required_token_positions_per_second": 28499.66984116284,
+      "required_pages_per_second_upper_estimate": 0.0017085116905015505
     },
     {
       "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": 29043.83681100077,
-      "required_pages_per_second_upper_estimate": 0.00225803056171825
+      "required_token_positions_per_second": 28499.66984116284,
+      "required_pages_per_second_upper_estimate": 0.0022157239733508944
     },
     {
       "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": 29043.83681100077,
-      "required_pages_per_second_upper_estimate": 0.37254315376920216
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28499.66984116284,
+      "required_pages_per_second_upper_estimate": 0.3655631641610913
     }
   ],
   "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": 64,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.09924472569522891,
-    "producer_p95_seconds": 0.2953243410011055,
-    "gpu_update_seconds": 0.9519403438297768,
-    "producer_batches_per_second": 10.076102210921562,
-    "gpu_batches_per_second": 1.050485995768257,
-    "producer_headroom": 9.591848203128645,
+    "producer_batch_seconds": 0.15374018804675416,
+    "producer_p95_seconds": 0.36335070499626454,
+    "gpu_update_seconds": 0.9701165014924927,
+    "producer_batches_per_second": 6.504480140845727,
+    "gpu_batches_per_second": 1.030804030713355,
+    "producer_headroom": 6.310103518264653,
     "sustainable": true,
-    "buffer_seconds": 60.92418200510571,
+    "buffer_seconds": 60.14722309253455,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 87,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 464883.5979835728,
-    "upload_bytes_per_second": 4976341.388513764,
+    "generation_bytes_per_second": 292008.66332315706,
+    "upload_bytes_per_second": 4983567.38343802,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 99.75319819097889,
+    "estimated_full_shards_per_hour": 62.65826153536829,
     "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": 1.124440429997776,
-      "bytes": 5889635,
-      "pages": 28,
+      "seconds": 0.9343277210027736,
+      "bytes": 6872118,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.0922325869978522,
-      "bytes": 6027814,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.3191045509993273,
+      "bytes": 10179142,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.0737022979992616,
-      "bytes": 6591396,
-      "pages": 31,
+      "seconds": 1.5255058740003733,
+      "bytes": 7690292,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1665754960013146,
-      "bytes": 11803743,
-      "pages": 35,
+      "seconds": 1.0536525450006593,
+      "bytes": 865655,
+      "pages": 27,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1389114759986114,
+      "bytes": 7332921,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.962029407000955,
-      "bytes": 2665696,
-      "pages": 38,
+      "seconds": 0.8940556989982724,
+      "bytes": 4992530,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.1741747699998086,
-      "bytes": 5919198,
-      "pages": 26,
-      "shards": 2
+      "seconds": 1.0987542440016114,
+      "bytes": 12106318,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.178343194998888,
-      "bytes": 10619543,
-      "pages": 42,
+      "seconds": 1.2125764899974456,
+      "bytes": 7340281,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.036487166002189,
-      "bytes": 2690894,
+      "seconds": 0.8264875540007779,
+      "bytes": 1223987,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2477019600009953,
+      "bytes": 12351474,
+      "pages": 36,
+      "shards": 5
+    },
+    {
+      "seconds": 1.4946348649973515,
+      "bytes": 6250315,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1432905639994715,
-      "bytes": 5745742,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.1129696229982073,
+      "bytes": 1097015,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.232796011998289,
-      "bytes": 13189628,
-      "pages": 37,
+      "seconds": 1.8527486320017488,
+      "bytes": 12857725,
+      "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 0.943955818998802,
-      "bytes": 5281939,
-      "pages": 35,
+      "seconds": 0.9267861950029328,
+      "bytes": 5234150,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.0029552209998656,
-      "bytes": 6176365,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.1693340630008606,
+      "bytes": 7789873,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.1109868420026032,
-      "bytes": 2592499,
-      "pages": 35,
+      "seconds": 1.150511206000374,
+      "bytes": 5980828,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.24596934100191,
-      "bytes": 5407802,
-      "pages": 21,
-      "shards": 2
+      "seconds": 0.8373393530018802,
+      "bytes": 1013057,
+      "pages": 28,
+      "shards": 1
     },
     {
-      "seconds": 1.361496919002093,
-      "bytes": 10822221,
+      "seconds": 1.2470868550008163,
+      "bytes": 11490913,
       "pages": 38,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.3625388020009268,
-      "bytes": 8506674,
-      "pages": 41,
+      "seconds": 13.443922794998798,
+      "bytes": 11916771,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.141177229997993,
-      "bytes": 6001543,
-      "pages": 29,
+      "seconds": 0.9173718489983003,
+      "bytes": 2617778,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9294679570011795,
-      "bytes": 2588442,
+      "seconds": 1.2762991070012504,
+      "bytes": 10864789,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.24567458599995,
+      "bytes": 1188999,
       "pages": 35,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.1605264330028149,
-      "bytes": 9613122,
-      "pages": 38,
-      "shards": 4
+      "seconds": 1.4343414930008294,
+      "bytes": 7437168,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.0401736989988422,
-      "bytes": 6081990,
+      "seconds": 1.4832610999983444,
+      "bytes": 5959171,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9755379390007874,
-      "bytes": 2340646,
-      "pages": 28,
+      "seconds": 0.9623825809976552,
+      "bytes": 6329110,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.4557110550013022,
-      "bytes": 16684165,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 1.2760817240014148,
-      "bytes": 7386323,
+      "seconds": 1.5357303359996877,
+      "bytes": 11699991,
       "pages": 39,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.0108026320012868,
-      "bytes": 815856,
-      "pages": 26,
+      "seconds": 1.0364311270022881,
+      "bytes": 982135,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 0.99785261299985,
-      "bytes": 326991,
-      "pages": 9,
-      "shards": 1
+      "seconds": 1.1774842960003298,
+      "bytes": 7504480,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.214567277002061,
-      "bytes": 5649266,
-      "pages": 28,
+      "seconds": 1.4250826019997476,
+      "bytes": 6016198,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2461040010020952,
-      "bytes": 11349081,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0920391100007691,
-      "bytes": 1249324,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.177016456000274,
+      "bytes": 5382045,
+      "pages": 34,
+      "shards": 2
     },
     {
-      "seconds": 1.1804772159994172,
-      "bytes": 13462589,
-      "pages": 36,
+      "seconds": 1.5666041279982892,
+      "bytes": 13422269,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 0.8390263229994162,
-      "bytes": 877453,
-      "pages": 26,
+      "seconds": 1.324093607996474,
+      "bytes": 1086878,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.1113367050020315,
-      "bytes": 10487773,
+      "seconds": 1.0723825640016003,
+      "bytes": 6205162,
       "pages": 35,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.1902512640008354,
-      "bytes": 7677939,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.3036935650015948,
+      "bytes": 11727766,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.0722998880009982,
-      "bytes": 987240,
-      "pages": 32,
+      "seconds": 0.7984904570039362,
+      "bytes": 1227271,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.0394909079986974,
-      "bytes": 6071704,
+      "seconds": 1.0684821660033776,
+      "bytes": 7548903,
       "pages": 32,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.4701623389992164,
-      "bytes": 17195357,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.2989959650003584,
+      "bytes": 11854087,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 0.8250811970028735,
-      "bytes": 1111187,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.0731057049997617,
+      "bytes": 5340693,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.2127139630028978,
-      "bytes": 7187251,
-      "pages": 30,
+      "seconds": 1.125931346003199,
+      "bytes": 7735546,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.2223667069993098,
-      "bytes": 5868379,
-      "pages": 38,
-      "shards": 2
+      "seconds": 0.949357431003591,
+      "bytes": 856694,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 0.9594618119990628,
-      "bytes": 5817766,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.058450859003642,
+      "bytes": 7478371,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 1.0792979020006896,
-      "bytes": 7541421,
-      "pages": 39,
+      "seconds": 1.0343096720025642,
+      "bytes": 11122347,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 0.9343277210027736,
-      "bytes": 6872118,
-      "pages": 34,
+      "seconds": 1.0296640300002764,
+      "bytes": 6697132,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3191045509993273,
-      "bytes": 10179142,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5255058740003733,
-      "bytes": 7690292,
+      "seconds": 1.1917317950064898,
+      "bytes": 7979357,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.0536525450006593,
-      "bytes": 865655,
+      "seconds": 0.9077869020038634,
+      "bytes": 931830,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.1389114759986114,
-      "bytes": 7332921,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.1493673139993916,
+      "bytes": 11660613,
+      "pages": 43,
+      "shards": 4
     },
     {
-      "seconds": 0.8940556989982724,
-      "bytes": 4992530,
-      "pages": 32,
+      "seconds": 1.1604952449997654,
+      "bytes": 5793744,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0987542440016114,
-      "bytes": 12106318,
-      "pages": 36,
-      "shards": 3
+      "seconds": 0.8524275599993416,
+      "bytes": 1090009,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.2125764899974456,
-      "bytes": 7340281,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.149581274999946,
+      "bytes": 13144315,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 0.8264875540007779,
-      "bytes": 1223987,
-      "pages": 36,
-      "shards": 1
+      "seconds": 1.5471423709968803,
+      "bytes": 10740667,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.2477019600009953,
-      "bytes": 12351474,
+      "seconds": 0.8488108419987839,
+      "bytes": 1104603,
       "pages": 36,
-      "shards": 5
+      "shards": 1
     },
     {
-      "seconds": 1.4946348649973515,
-      "bytes": 6250315,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.9554288359940983,
+      "bytes": 7500029,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.1129696229982073,
-      "bytes": 1097015,
-      "pages": 34,
+      "seconds": 0.8447882219988969,
+      "bytes": 1059224,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.8527486320017488,
-      "bytes": 12857725,
-      "pages": 32,
+      "seconds": 1.164817237004172,
+      "bytes": 10956336,
+      "pages": 35,
       "shards": 4
     },
     {
-      "seconds": 0.9267861950029328,
-      "bytes": 5234150,
-      "pages": 32,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1693340630008606,
-      "bytes": 7789873,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 1.150511206000374,
-      "bytes": 5980828,
+      "seconds": 1.2077032870001858,
+      "bytes": 11770097,
       "pages": 34,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 0.8373393530018802,
-      "bytes": 1013057,
-      "pages": 28,
+      "seconds": 0.8165655029952177,
+      "bytes": 1121083,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.2470868550008163,
-      "bytes": 11490913,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 13.443922794998798,
-      "bytes": 11916771,
-      "pages": 38,
+      "seconds": 0.944058788001712,
+      "bytes": 7283383,
+      "pages": 27,
       "shards": 3
     },
     {
-      "seconds": 0.9173718489983003,
-      "bytes": 2617778,
+      "seconds": 1.194725619003293,
+      "bytes": 5214434,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2762991070012504,
-      "bytes": 10864789,
-      "pages": 36,
+      "seconds": 1.2110685700026806,
+      "bytes": 6209123,
+      "pages": 38,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3596407239965629,
+      "bytes": 7496236,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.24567458599995,
-      "bytes": 1188999,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.0467049810031313,
+      "bytes": 6666834,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.4343414930008294,
-      "bytes": 7437168,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.164912510997965,
+      "bytes": 12226977,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.4832610999983444,
-      "bytes": 5959171,
-      "pages": 37,
+      "seconds": 1.2401886210063822,
+      "bytes": 6134452,
+      "pages": 38,
       "shards": 2
+    },
+    {
+      "seconds": 1.449164785000903,
+      "bytes": 969493,
+      "pages": 29,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e6a73d2db07c024ebb237c7d0b6fcb773796b3b6ab1195a52d429d55bc86e73a
-size 721766911
+oid sha256:04c2d546da542c12c58a0435f1ba889816d6f1028daf377c51a35cccae028102
+size 721764095

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 64,
-  "prefetch_low_water": 32,
+  "prefetch_batches": 62,
+  "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,