Qwen90M step 745000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 740000,
+  "step": 745000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9",
+    "model.safetensors": "0d7ce863025d2619d914fb70f1a1e4a1068b7ce5609365ceb4f22dff6ccfaf85",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393",
-    "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037",
+    "trainer.pt": "d44898a301bc80cd4b55a3e048afe4856a384a2f161cf4889c3ae4e68e12da5a",
+    "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687",
     "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379",
-    "shard_scheduler.json": "7f8712a7f9538f373e624550e1f5a9ce3a8e415d6a140e719f8256eddac3ad48",
-    "coverage.json": "7585df2fd9672883a1500e76331661f838d543263097041bee9efaf33517c187",
+    "shard_scheduler.json": "b7c0bd5b9394f7c75d2d94d6f2fc1c2fae65583ba811d4b0353f995a152acb08",
+    "coverage.json": "798d5521406ab366e17be4a5abaa9337e274ccadae7107b1abfdcd9cbdebec16",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 740000,
+  "step": 745000,
   "stats": {
-    "climbmix/blocks": 26640000,
-    "climbmix/targets": 44534380574,
-    "climbmix/input_tokens": 44537202395,
-    "ultra_qa/blocks": 9990000,
-    "ultra_qa/targets": 16297612094,
-    "ultra_qa/input_tokens": 16297643521,
-    "cortex/blocks": 9990000,
-    "cortex/targets": 379836958,
-    "cortex/input_tokens": 15879487851,
-    "ultra_style/blocks": 9990000,
-    "ultra_style/targets": 15801366047,
-    "ultra_style/input_tokens": 15802194940,
-    "openmath/blocks": 6660000,
-    "openmath/targets": 10291092474,
-    "openmath/input_tokens": 10294521411,
-    "rewrite6/blocks": 3330000,
-    "rewrite6/targets": 5459078580,
-    "rewrite6/input_tokens": 5465063978
+    "climbmix/blocks": 26820000,
+    "climbmix/targets": 44835016967,
+    "climbmix/input_tokens": 44837857904,
+    "ultra_qa/blocks": 10057500,
+    "ultra_qa/targets": 16407642063,
+    "ultra_qa/input_tokens": 16407673700,
+    "cortex/blocks": 10057500,
+    "cortex/targets": 382400656,
+    "cortex/input_tokens": 15986736473,
+    "ultra_style/blocks": 10057500,
+    "ultra_style/targets": 15908518038,
+    "ultra_style/input_tokens": 15909352153,
+    "openmath/blocks": 6705000,
+    "openmath/targets": 10360628323,
+    "openmath/input_tokens": 10364080425,
+    "rewrite6/blocks": 3352500,
+    "rewrite6/targets": 5495963673,
+    "rewrite6/input_tokens": 5501989340
   },
   "sources": {
     "climbmix": {
-      "page": 3648,
-      "row": 803,
+      "page": 3672,
+      "row": 3991,
       "start": {
-        "file": 176,
+        "file": 177,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 48,
+          "row_group": 62,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 14943011
+      "documents": 15044503
     },
     "rewrite6": {
-      "page": 25,
-      "row": 3061,
+      "page": 39,
+      "row": 3164,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 102400
+          "row": 159744
         }
       },
       "epoch": 41,
-      "offset": 2048,
-      "documents": 8527025
+      "offset": 0,
+      "documents": 8584472
     },
     "openmath": {
-      "page": 2280,
-      "row": 890,
+      "page": 2295,
+      "row": 2630,
       "start": {
         "file": 21,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 169,
-          "row": 230
+          "row_group": 230,
+          "row": 670
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 9339770
+      "documents": 9402950
     },
     "ultra_qa": {
-      "page": 985,
-      "row": 3725,
+      "page": 992,
+      "row": 2453,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 9755
+          "row": 38427
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 4038285
+      "offset": 4096,
+      "documents": 4065685
     },
     "ultra_style": {
-      "page": 1504,
-      "row": 3284,
+      "page": 1514,
+      "row": 3015,
       "start": {
-        "file": 8,
+        "file": 9,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 84737
+          "row_group": 0,
+          "row": 36870
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 6163668
+      "documents": 6204359
     },
     "cortex": {
-      "page": 193744,
-      "row": 26,
+      "page": 195052,
+      "row": 49,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 24799258
+      "documents": 24966705
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 740000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393", "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "7f8712a7f9538f373e624550e1f5a9ce3a8e415d6a140e719f8256eddac3ad48", "coverage.json": "7585df2fd9672883a1500e76331661f838d543263097041bee9efaf33517c187", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000740000"}
+{"step": 745000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "0d7ce863025d2619d914fb70f1a1e4a1068b7ce5609365ceb4f22dff6ccfaf85", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "d44898a301bc80cd4b55a3e048afe4856a384a2f161cf4889c3ae4e68e12da5a", "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "b7c0bd5b9394f7c75d2d94d6f2fc1c2fae65583ba811d4b0353f995a152acb08", "coverage.json": "798d5521406ab366e17be4a5abaa9337e274ccadae7107b1abfdcd9cbdebec16", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000745000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9
+oid sha256:0d7ce863025d2619d914fb70f1a1e4a1068b7ce5609365ceb4f22dff6ccfaf85
 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": 75740.10139994962,
-      "required_pages_per_second_upper_estimate": 0.005239875956096181
+      "required_token_positions_per_second": 76481.61050310897,
+      "required_pages_per_second_upper_estimate": 0.005291175276390908
     },
     {
       "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": 9467.512674993703,
-      "required_pages_per_second_upper_estimate": 0.003963849852873521
+      "required_token_positions_per_second": 9560.201312888621,
+      "required_pages_per_second_upper_estimate": 0.004002656650001265
     },
     {
       "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": 18935.025349987405,
-      "required_pages_per_second_upper_estimate": 0.0039436664302040505
+      "required_token_positions_per_second": 19120.402625777242,
+      "required_pages_per_second_upper_estimate": 0.003982275628023559
     },
     {
       "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": 28402.538024981106,
-      "required_pages_per_second_upper_estimate": 0.0017026887864331514
+      "required_token_positions_per_second": 28680.603938665863,
+      "required_pages_per_second_upper_estimate": 0.0017193584133764865
     },
     {
       "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": 28402.538024981106,
-      "required_pages_per_second_upper_estimate": 0.002208172401880472
+      "required_token_positions_per_second": 28680.603938665863,
+      "required_pages_per_second_upper_estimate": 0.002229790803586767
     },
     {
       "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": 22,
-      "required_token_positions_per_second": 28402.538024981106,
-      "required_pages_per_second_upper_estimate": 0.3643172615151307
+      "estimated_pages_for_buffer": 23,
+      "required_token_positions_per_second": 28680.603938665863,
+      "required_pages_per_second_upper_estimate": 0.3678839924919622
     }
   ],
   "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": 62,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.11142892917197855,
-    "producer_p95_seconds": 0.2649414719999186,
-    "gpu_update_seconds": 0.973434133797569,
-    "producer_batches_per_second": 8.974330162112638,
-    "gpu_batches_per_second": 1.0272908718526153,
-    "producer_headroom": 8.735919307769514,
+    "producer_batch_seconds": 0.07352219521106917,
+    "producer_p95_seconds": 0.255219880000368,
+    "gpu_update_seconds": 0.9639964367251781,
+    "producer_batches_per_second": 13.601334904775047,
+    "gpu_batches_per_second": 1.0373482327353105,
+    "producer_headroom": 13.111638382908936,
     "sustainable": true,
-    "buffer_seconds": 60.35291629544928,
+    "buffer_seconds": 60.73177551368622,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 287306.2932266875,
-    "upload_bytes_per_second": 5098154.501941984,
+    "generation_bytes_per_second": 392318.91666500096,
+    "upload_bytes_per_second": 5364724.667990984,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 61.64924237823934,
+    "estimated_full_shards_per_hour": 84.18250679934046,
     "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": 1.2874604959997669,
-      "bytes": 2530940,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.3973069029998442,
+      "bytes": 13145960,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.3693337599997903,
-      "bytes": 10798536,
-      "pages": 37,
+      "seconds": 1.3652075969985162,
+      "bytes": 5700613,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.092897812000956,
-      "bytes": 6537422,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.7661054199998034,
+      "bytes": 7454533,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9842067560020951,
+      "bytes": 1088907,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 0.9582660920004855,
-      "bytes": 2603502,
+      "seconds": 1.1375966550003795,
+      "bytes": 6241738,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5727497110001423,
-      "bytes": 11757823,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.329601728997659,
+      "bytes": 12006565,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1941701559990179,
-      "bytes": 5221736,
-      "pages": 34,
+      "seconds": 1.637713142998109,
+      "bytes": 6812710,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.6056268569991516,
-      "bytes": 7578967,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.20693444599965,
-      "bytes": 6453330,
-      "pages": 35,
+      "seconds": 1.216705439001089,
+      "bytes": 6221386,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9651811909989192,
-      "bytes": 2682508,
-      "pages": 37,
+      "seconds": 1.2352482690002944,
+      "bytes": 2570616,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.4213833419998991,
-      "bytes": 10754179,
-      "pages": 34,
+      "seconds": 1.360712293000688,
+      "bytes": 10168781,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.5129684479998105,
-      "bytes": 6722253,
+      "seconds": 1.3043567909990088,
+      "bytes": 7359704,
       "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.165229547999843,
-      "bytes": 7984438,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.5744073349997052,
+      "bytes": 1306429,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.001051560999258,
-      "bytes": 939443,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.5331175120008993,
+      "bytes": 11738494,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.4753997340012575,
-      "bytes": 11016252,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0237298360007117,
+      "bytes": 7075351,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.6974678700007644,
-      "bytes": 6266676,
-      "pages": 36,
+      "seconds": 1.4313278709996666,
+      "bytes": 6586683,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.6492898739998054,
-      "bytes": 1094803,
-      "pages": 34,
+      "seconds": 1.1789775520010153,
+      "bytes": 937197,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.8326161909990333,
-      "bytes": 13393499,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0937288770001032,
+      "bytes": 7693834,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.4041516550005326,
-      "bytes": 11034703,
-      "pages": 37,
+      "seconds": 1.3860260590008693,
+      "bytes": 10598757,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1081106500005262,
-      "bytes": 1056276,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.0763569690025179,
+      "bytes": 2341572,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.211192590999417,
-      "bytes": 7370118,
-      "pages": 35,
+      "seconds": 1.3760699319973355,
+      "bytes": 11906705,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.246054737999657,
-      "bytes": 1155172,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2023094430005585,
+      "bytes": 5878390,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.6413428979994933,
-      "bytes": 11205821,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.9879657829987991,
+      "bytes": 6822345,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.3393069079993438,
-      "bytes": 11931034,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.0887722430015856,
+      "bytes": 6046508,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9887657250001212,
-      "bytes": 1035853,
-      "pages": 34,
+      "seconds": 0.9664684920026048,
+      "bytes": 1011694,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.3170902919991931,
-      "bytes": 7309091,
-      "pages": 30,
+      "seconds": 1.350358371000766,
+      "bytes": 7369637,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.3663103399994725,
-      "bytes": 5168575,
-      "pages": 32,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1543622389999655,
-      "bytes": 6027007,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.1391658569991705,
+      "bytes": 11247930,
+      "pages": 27,
+      "shards": 3
     },
     {
-      "seconds": 1.39643492399955,
-      "bytes": 7665740,
+      "seconds": 1.2541976089996751,
+      "bytes": 7910419,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1344610849992023,
-      "bytes": 6680190,
-      "pages": 29,
+      "seconds": 1.501875165002275,
+      "bytes": 6144552,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.5331677430003765,
-      "bytes": 11639669,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.1508353210010682,
+      "bytes": 1193747,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 0.9482507379998424,
-      "bytes": 1158257,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.6794259740017878,
+      "bytes": 11268683,
+      "pages": 32,
+      "shards": 4
     },
     {
-      "seconds": 1.2962373409991415,
-      "bytes": 6282079,
+      "seconds": 1.0690768630011007,
+      "bytes": 5982757,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.4015076579999004,
-      "bytes": 7431338,
-      "pages": 31,
+      "seconds": 1.3195378630007326,
+      "bytes": 8379086,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.9902176969990251,
-      "bytes": 5564439,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.5084312399994815,
+      "bytes": 5915466,
+      "pages": 28,
+      "shards": 3
     },
     {
-      "seconds": 1.667742324003484,
-      "bytes": 13340513,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.1589784089992463,
+      "bytes": 5695458,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.2163241669986746,
-      "bytes": 1010699,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.9208255039993674,
+      "bytes": 7437563,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.3674872539995704,
-      "bytes": 6041473,
+      "seconds": 1.6179820170000312,
+      "bytes": 6082981,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3613106869997864,
-      "bytes": 11006649,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.991337299998122,
+      "bytes": 1324376,
+      "pages": 36,
+      "shards": 1
     },
     {
-      "seconds": 0.8713928039978782,
-      "bytes": 1169048,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.2612079769969569,
+      "bytes": 16881065,
+      "pages": 36,
+      "shards": 5
     },
     {
-      "seconds": 1.0586420059989905,
-      "bytes": 6189631,
-      "pages": 35,
+      "seconds": 1.1940166240019607,
+      "bytes": 6068573,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.3973069029998442,
-      "bytes": 13145960,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0544981790008023,
+      "bytes": 2562621,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3652075969985162,
-      "bytes": 5700613,
+      "seconds": 1.3444492599992373,
+      "bytes": 6214787,
       "pages": 36,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.7661054199998034,
-      "bytes": 7454533,
-      "pages": 30,
+      "seconds": 1.2648084209977242,
+      "bytes": 5986982,
+      "pages": 31,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4115151900005003,
+      "bytes": 7731251,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9842067560020951,
-      "bytes": 1088907,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1964002810018428,
+      "bytes": 11692696,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 1.1375966550003795,
-      "bytes": 6241738,
-      "pages": 36,
+      "seconds": 1.0258839710004395,
+      "bytes": 2700953,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.329601728997659,
-      "bytes": 12006565,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.2136837799989735,
+      "bytes": 9835275,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.637713142998109,
-      "bytes": 6812710,
-      "pages": 30,
+      "seconds": 1.2959440570011793,
+      "bytes": 5942943,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.216705439001089,
-      "bytes": 6221386,
-      "pages": 36,
+      "seconds": 0.9916904240017175,
+      "bytes": 2484673,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2352482690002944,
-      "bytes": 2570616,
-      "pages": 35,
+      "seconds": 1.0907142409996595,
+      "bytes": 6115368,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.360712293000688,
-      "bytes": 10168781,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.3705160650024482,
+      "bytes": 13033658,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 1.3043567909990088,
-      "bytes": 7359704,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.2911444290002692,
+      "bytes": 5949225,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.5744073349997052,
-      "bytes": 1306429,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.0684557280001172,
+      "bytes": 6155327,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.5331175120008993,
-      "bytes": 11738494,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.2350120470000547,
+      "bytes": 2680827,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0237298360007117,
-      "bytes": 7075351,
-      "pages": 32,
+      "seconds": 1.5306445749993145,
+      "bytes": 9835087,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.4313278709996666,
-      "bytes": 6586683,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.9580864189993008,
+      "bytes": 989214,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.1789775520010153,
-      "bytes": 937197,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.3733238169988908,
+      "bytes": 13560061,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.0937288770001032,
-      "bytes": 7693834,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.2763601959995867,
+      "bytes": 5814992,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.3860260590008693,
-      "bytes": 10598757,
-      "pages": 35,
+      "seconds": 1.203305928996997,
+      "bytes": 7121753,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.0763569690025179,
-      "bytes": 2341572,
-      "pages": 29,
+      "seconds": 1.1241114900003595,
+      "bytes": 6180535,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3760699319973355,
-      "bytes": 11906705,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.1907369090004067,
+      "bytes": 6141214,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.2023094430005585,
-      "bytes": 5878390,
+      "seconds": 0.9464336839992029,
+      "bytes": 2380354,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.9879657829987991,
-      "bytes": 6822345,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.6319502810001723,
+      "bytes": 16693932,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.0887722430015856,
-      "bytes": 6046508,
+      "seconds": 0.91435248900234,
+      "bytes": 2656824,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9664684920026048,
-      "bytes": 1011694,
-      "pages": 29,
-      "shards": 1
+      "seconds": 1.1157520460001251,
+      "bytes": 5879867,
+      "pages": 28,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393
-size 721766975
+oid sha256:d44898a301bc80cd4b55a3e048afe4856a384a2f161cf4889c3ae4e68e12da5a
+size 721775039

training_config.json

@@ -53,7 +53,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 62,
+  "prefetch_batches": 63,
   "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,