Qwen90M step 485000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 483230,
+  "step": 485000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "3091ee2c20e9e7b92f2fbd1e0f7e26d27ee1f69d13a57fb3cec81e93a601cebe",
+    "model.safetensors": "7fd364f1b254abb0a762a84c0a7c714c534994ffdbdf3dc30a0b7cae01ffeb6e",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "dce3605b8fce93509920de867adac7256115c7fda251e6af444df4ef1a81445a",
-    "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
-    "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
-    "shard_scheduler.json": "39a4c183a128fcd08cafa3c3e7c2cdf0b813ba4d0d8bcde2764b7a2fb143e4f4",
-    "coverage.json": "04878bda51568cbea45cc34a1b0c8774e481860c8887b6161ce50e0119dd075f",
+    "trainer.pt": "2e31b934cad9f27085ea79b3fd3db6d8ab83a24db1f9125cd66705d726ad26e8",
+    "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
+    "performance.json": "e2fbfc542f997748539ec27cc61ae6365d39122628b292c00db059dd801c9082",
+    "shard_scheduler.json": "b8e4654ade0997d04db51bcd77ba9d9a995dfd09656a2adb9fbb8efc28315f5f",
+    "coverage.json": "32c5b1549b70f2106c3b9addeed59093c0fe5ba7d9a8bd64ba52215c1518731c",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 483230,
+  "step": 485000,
   "stats": {
-    "climbmix/blocks": 17396280,
-    "climbmix/targets": 29078088464,
-    "climbmix/input_tokens": 29079931166,
-    "ultra_qa/blocks": 6523605,
-    "ultra_qa/targets": 10642172522,
-    "ultra_qa/input_tokens": 10642193016,
-    "cortex/blocks": 6523605,
-    "cortex/targets": 248114285,
-    "cortex/input_tokens": 10370213106,
-    "ultra_style/blocks": 6523605,
-    "ultra_style/targets": 10318245277,
-    "ultra_style/input_tokens": 10318785974,
-    "openmath/blocks": 4349070,
-    "openmath/targets": 6720156497,
-    "openmath/input_tokens": 6722395717,
-    "rewrite6/blocks": 2174535,
-    "rewrite6/targets": 3564874771,
-    "rewrite6/input_tokens": 3568783073
+    "climbmix/blocks": 17460000,
+    "climbmix/targets": 29184768142,
+    "climbmix/input_tokens": 29186617700,
+    "ultra_qa/blocks": 6547500,
+    "ultra_qa/targets": 10681167812,
+    "ultra_qa/input_tokens": 10681188377,
+    "cortex/blocks": 6547500,
+    "cortex/targets": 249030011,
+    "cortex/input_tokens": 10408163590,
+    "ultra_style/blocks": 6547500,
+    "ultra_style/targets": 10356067453,
+    "ultra_style/input_tokens": 10356610174,
+    "openmath/blocks": 4365000,
+    "openmath/targets": 6744753340,
+    "openmath/input_tokens": 6747000681,
+    "rewrite6/blocks": 2182500,
+    "rewrite6/targets": 3577914913,
+    "rewrite6/input_tokens": 3581837409
   },
   "sources": {
     "climbmix": {
-      "page": 2383,
-      "row": 126,
+      "page": 2391,
+      "row": 3140,
       "start": {
         "file": 115,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 26,
+          "row_group": 58,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 9760894
+      "documents": 9796676
     },
     "rewrite6": {
-      "page": 5,
-      "row": 1536,
+      "page": 10,
+      "row": 1343,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 20480
+          "row": 40960
         }
       },
       "epoch": 27,
       "offset": 0,
-      "documents": 5567924
+      "documents": 5588211
     },
     "openmath": {
-      "page": 1489,
-      "row": 259,
+      "page": 1494,
+      "row": 2026,
       "start": {
-        "file": 13,
+        "file": 14,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 422,
-          "row": 494
+          "row_group": 6,
+          "row": 324
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 6099203
+      "documents": 6121450
     },
     "ultra_qa": {
-      "page": 643,
-      "row": 3346,
+      "page": 646,
+      "row": 785,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 35599
+          "row": 47887
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 2637074
+      "offset": 2048,
+      "documents": 2646801
     },
     "ultra_style": {
-      "page": 982,
-      "row": 1968,
+      "page": 986,
+      "row": 411,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 1917
+          "row": 18301
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4024240
+      "documents": 4039067
     },
     "cortex": {
-      "page": 126525,
-      "row": 34,
+      "page": 126988,
+      "row": 6,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 16195234
+      "documents": 16254470
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 483230, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3091ee2c20e9e7b92f2fbd1e0f7e26d27ee1f69d13a57fb3cec81e93a601cebe", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "dce3605b8fce93509920de867adac7256115c7fda251e6af444df4ef1a81445a", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "39a4c183a128fcd08cafa3c3e7c2cdf0b813ba4d0d8bcde2764b7a2fb143e4f4", "coverage.json": "04878bda51568cbea45cc34a1b0c8774e481860c8887b6161ce50e0119dd075f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000483230"}
+{"step": 485000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "7fd364f1b254abb0a762a84c0a7c714c534994ffdbdf3dc30a0b7cae01ffeb6e", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2e31b934cad9f27085ea79b3fd3db6d8ab83a24db1f9125cd66705d726ad26e8", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "e2fbfc542f997748539ec27cc61ae6365d39122628b292c00db059dd801c9082", "shard_scheduler.json": "b8e4654ade0997d04db51bcd77ba9d9a995dfd09656a2adb9fbb8efc28315f5f", "coverage.json": "32c5b1549b70f2106c3b9addeed59093c0fe5ba7d9a8bd64ba52215c1518731c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000485000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3091ee2c20e9e7b92f2fbd1e0f7e26d27ee1f69d13a57fb3cec81e93a601cebe
+oid sha256:7fd364f1b254abb0a762a84c0a7c714c534994ffdbdf3dc30a0b7cae01ffeb6e
 size 360622744

performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 90,
     "mode": "compiled",
-    "microbatch_seconds": 0.9383984099999907,
+    "microbatch_seconds": 0.9398671289999925,
     "peak_gb": 89.37790203094482,
     "fits_budget": true,
-    "estimated_update_seconds": 0.9383984099999907,
-    "input_tokens_per_second": 196419.77014858948
+    "estimated_update_seconds": 0.9398671289999925,
+    "input_tokens_per_second": 196112.82734838733
   },
   "measurements": [
     {

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": 76395.03562198069,
-      "required_pages_per_second_upper_estimate": 0.005285185825233053
+      "required_token_positions_per_second": 75745.19625580676,
+      "required_pages_per_second_upper_estimate": 0.005240228430046073
     },
     {
       "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": 9549.379452747586,
-      "required_pages_per_second_upper_estimate": 0.003998125763146351
+      "required_token_positions_per_second": 9468.149531975845,
+      "required_pages_per_second_upper_estimate": 0.003964116491592859
     },
     {
       "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": 19098.75890549517,
-      "required_pages_per_second_upper_estimate": 0.0039777678118720905
+      "required_token_positions_per_second": 18936.29906395169,
+      "required_pages_per_second_upper_estimate": 0.003943931711232715
     },
     {
       "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": 28648.138358242755,
-      "required_pages_per_second_upper_estimate": 0.0017174121514021922
+      "required_token_positions_per_second": 28404.448595927533,
+      "required_pages_per_second_upper_estimate": 0.0017028033222441167
     },
     {
       "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": 28648.138358242755,
-      "required_pages_per_second_upper_estimate": 0.0022272667475098673
+      "required_token_positions_per_second": 28404.448595927533,
+      "required_pages_per_second_upper_estimate": 0.0022083209403678437
     },
     {
       "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": 28648.138358242755,
-      "required_pages_per_second_upper_estimate": 0.3674675588851189
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28404.448595927533,
+      "required_pages_per_second_upper_estimate": 0.3643417682678202
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,412 +155,100 @@
   "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": 0.13329735316432334,
-    "producer_p95_seconds": 0.36278332399524515,
-    "gpu_update_seconds": 0.9650888882992639,
-    "producer_batches_per_second": 7.50202443080203,
-    "gpu_batches_per_second": 1.0361739857582015,
-    "producer_headroom": 7.240120417916649,
+    "producer_batch_seconds": 0.16442591646878846,
+    "producer_p95_seconds": 0.13880198200013183,
+    "gpu_update_seconds": 0.973368657611048,
+    "producer_batches_per_second": 6.081766314435117,
+    "gpu_batches_per_second": 1.0273599752577955,
+    "producer_headroom": 5.919800713385801,
     "sustainable": true,
-    "buffer_seconds": 60.800599962853624,
+    "buffer_seconds": 60.348856771884975,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 69,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 298978.01718284166,
-    "upload_bytes_per_second": 5442207.661463906,
+    "generation_bytes_per_second": 424533.6472953534,
+    "upload_bytes_per_second": 5696740.445231295,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 64.15372263540208,
+    "estimated_full_shards_per_hour": 91.09503807206585,
     "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": 2.2645728169954964,
-      "bytes": 6235357,
-      "pages": 44,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3341411860019434,
-      "bytes": 11745696,
-      "pages": 39,
-      "shards": 4
-    },
-    {
-      "seconds": 3.1945914269963396,
-      "bytes": 906980,
-      "pages": 27,
+      "seconds": 0.9475190029999112,
+      "bytes": 21104,
+      "pages": 1,
       "shards": 1
     },
     {
-      "seconds": 1.100945904996479,
-      "bytes": 6366458,
+      "seconds": 1.029576297999938,
+      "bytes": 6528629,
       "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1544355079968227,
-      "bytes": 7474922,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0820681480036,
-      "bytes": 10987528,
-      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.2853161590028321,
-      "bytes": 6397328,
-      "pages": 37,
+      "seconds": 1.0352755850000221,
+      "bytes": 6365146,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.1332088359995396,
-      "bytes": 7397305,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 0.7730173130039475,
-      "bytes": 1053057,
-      "pages": 32,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3466375830030302,
-      "bytes": 11763121,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.1308769879999545,
+      "bytes": 6215001,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.4802930930018192,
-      "bytes": 1138429,
+      "seconds": 1.016927755999859,
+      "bytes": 7382400,
       "pages": 32,
-      "shards": 1
-    },
-    {
-      "seconds": 1.8897758500024793,
-      "bytes": 11833830,
-      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.1506007540010614,
-      "bytes": 2440099,
-      "pages": 31,
+      "seconds": 1.1139478459999737,
+      "bytes": 6306186,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.5953800970019074,
-      "bytes": 10985431,
-      "pages": 43,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5438401100036572,
-      "bytes": 7453011,
+      "seconds": 1.1660128299999997,
+      "bytes": 8072363,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0392796939995605,
-      "bytes": 1282728,
+      "seconds": 0.9177546119999533,
+      "bytes": 6210206,
       "pages": 37,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3358671570022125,
-      "bytes": 5881131,
-      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.5660001680007554,
-      "bytes": 16980575,
-      "pages": 44,
-      "shards": 6
-    },
-    {
-      "seconds": 0.8579309720007586,
-      "bytes": 927462,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3555858160034404,
-      "bytes": 7535281,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1513379759999225,
-      "bytes": 10862357,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9964048779947916,
-      "bytes": 1334873,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2506730089953635,
-      "bytes": 7336892,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3737510540013318,
-      "bytes": 6860686,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2590301210002508,
-      "bytes": 10720957,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 1.187173960999644,
-      "bytes": 7325771,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8256457960014814,
-      "bytes": 1213741,
+      "seconds": 1.1674345439998888,
+      "bytes": 5990051,
       "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0744473489976372,
-      "bytes": 7318826,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3007696439963183,
-      "bytes": 9895316,
-      "pages": 41,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3112097839984926,
-      "bytes": 6821672,
-      "pages": 34,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0360545559960883,
-      "bytes": 7415725,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9522623859957093,
-      "bytes": 963026,
-      "pages": 30,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1533947639982216,
-      "bytes": 11916585,
-      "pages": 41,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1312003819984966,
-      "bytes": 5813649,
-      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.0730638719978742,
-      "bytes": 1187099,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1143769740010612,
-      "bytes": 13033863,
-      "pages": 32,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1261608329950832,
-      "bytes": 10613783,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0124226469997666,
-      "bytes": 904800,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1419830579980044,
-      "bytes": 7590743,
+      "seconds": 1.0493045279999933,
+      "bytes": 6091029,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.8363852010006667,
-      "bytes": 1241882,
+      "seconds": 1.0872325649997947,
+      "bytes": 6195111,
       "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.234218167999643,
-      "bytes": 11088956,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0702078030008124,
-      "bytes": 12031494,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9152153160030139,
-      "bytes": 886815,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3266999849947751,
-      "bytes": 11836619,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9408924880044651,
-      "bytes": 1241676,
-      "pages": 37,
-      "shards": 1
-    },
-    {
-      "seconds": 1.40072601299471,
-      "bytes": 6187128,
-      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.231672369001899,
-      "bytes": 7550882,
-      "pages": 41,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3564282489969628,
-      "bytes": 10548156,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1907792190031614,
-      "bytes": 7251298,
-      "pages": 32,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9704950960003771,
-      "bytes": 1288101,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2859423129993957,
-      "bytes": 5955041,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.6710909179964801,
-      "bytes": 11523305,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8526758169973618,
-      "bytes": 1207490,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3300546199970995,
-      "bytes": 11866142,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8256019309992553,
-      "bytes": 2326710,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1752165859943489,
-      "bytes": 9508137,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.008012272999622,
-      "bytes": 7606530,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8981697260023793,
-      "bytes": 1206792,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.5740248830043129,
-      "bytes": 5500237,
-      "pages": 20,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1595400419973885,
-      "bytes": 7406603,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1209028549928917,
-      "bytes": 11633071,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5646861489949515,
-      "bytes": 6165459,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9572175139983301,
-      "bytes": 2339484,
-      "pages": 32,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3525367339971126,
-      "bytes": 9731842,
-      "pages": 31,
+      "seconds": 1.1556296060000477,
+      "bytes": 7640700,
+      "pages": 37,
       "shards": 3
     }
   ],
@@ -606,7 +294,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    2.3715912330000037,
-    0.09919368099997428
+    2.220795152999983,
+    0.01521137299999964
   ]
 }

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dce3605b8fce93509920de867adac7256115c7fda251e6af444df4ef1a81445a
-size 721773887
+oid sha256:2e31b934cad9f27085ea79b3fd3db6d8ab83a24db1f9125cd66705d726ad26e8
+size 721772095

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,