Qwen90M step 825000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 820000,
+  "step": 825000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "be65b8dd85d32e3c046a0a29d18dd9c9636bd3518c6180704165ccc3b345f97a",
+    "model.safetensors": "c94a248685b839a2edfdd0022e869649e1d13c7266eb40bee2cd5b65ffe131e7",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "72a6fa391959ac41be57caac41e5d9e33fa242f668e7e959e3b49726861e8403",
-    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
+    "trainer.pt": "de221e9ed21e83c30b24525b864378be67505342cacb8dfb75a9ebfcd0672c2b",
+    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
     "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
-    "shard_scheduler.json": "cca80237086820dcf5c494e00db56e92fb63f5bff088142a981f0d3db418bb87",
-    "coverage.json": "c13407e1cced45349ea4cc9f52ed38e019bd1fa9d3e900afc3652ad06d436a2c",
+    "shard_scheduler.json": "3a5912414de65c80e1ba8c4b974f6204935ab0245d673f1014baff16d50c52ea",
+    "coverage.json": "384700a0a13bd6ae2e77220c1d1c48f9e386f807b0b80377087bdf716560b06c",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 820000,
+  "step": 825000,
   "stats": {
-    "climbmix/blocks": 29520000,
-    "climbmix/targets": 49348064056,
-    "climbmix/input_tokens": 49351191260,
-    "ultra_qa/blocks": 11070000,
-    "ultra_qa/targets": 18059605638,
-    "ultra_qa/input_tokens": 18059640442,
-    "cortex/blocks": 11070000,
-    "cortex/targets": 420935370,
-    "cortex/input_tokens": 17596608233,
-    "ultra_style/blocks": 11070000,
-    "ultra_style/targets": 17509519663,
-    "ultra_style/input_tokens": 17510438016,
-    "openmath/blocks": 7380000,
-    "openmath/targets": 11403359242,
-    "openmath/input_tokens": 11407158801,
-    "rewrite6/blocks": 3690000,
-    "rewrite6/targets": 6049287294,
-    "rewrite6/input_tokens": 6055919347
+    "climbmix/blocks": 29700000,
+    "climbmix/targets": 49648833395,
+    "climbmix/input_tokens": 49651979405,
+    "ultra_qa/blocks": 11137500,
+    "ultra_qa/targets": 18169783675,
+    "ultra_qa/input_tokens": 18169818678,
+    "cortex/blocks": 11137500,
+    "cortex/targets": 423506385,
+    "cortex/input_tokens": 17703865080,
+    "ultra_style/blocks": 11137500,
+    "ultra_style/targets": 17616349514,
+    "ultra_style/input_tokens": 17617273526,
+    "openmath/blocks": 7425000,
+    "openmath/targets": 11472953325,
+    "openmath/input_tokens": 11476776268,
+    "rewrite6/blocks": 3712500,
+    "rewrite6/targets": 6086134136,
+    "rewrite6/input_tokens": 6092806682
   },
   "sources": {
     "climbmix": {
-      "page": 4042,
-      "row": 3493,
+      "page": 4067,
+      "row": 1808,
       "start": {
-        "file": 195,
+        "file": 196,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 54,
+          "row_group": 71,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 16559525
+      "offset": 4096,
+      "documents": 16660240
     },
     "rewrite6": {
-      "page": 50,
-      "row": 410,
+      "page": 14,
+      "row": 93,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 204800
+          "row": 57344
         }
       },
-      "epoch": 45,
+      "epoch": 46,
       "offset": 0,
-      "documents": 9448390
+      "documents": 9506021
     },
     "openmath": {
-      "page": 2526,
-      "row": 3013,
+      "page": 2542,
+      "row": 913,
       "start": {
         "file": 23,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 303,
-          "row": 546
+          "row_group": 369,
+          "row": 82
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 10349509
+      "documents": 10412945
     },
     "ultra_qa": {
-      "page": 1092,
-      "row": 2051,
+      "page": 1099,
+      "row": 700,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
-          "row": 57600
+          "row_group": 5,
+          "row": 21781
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4474883
+      "documents": 4502204
     },
     "ultra_style": {
-      "page": 1667,
-      "row": 1796,
+      "page": 1677,
+      "row": 2616,
       "start": {
-        "file": 9,
+        "file": 10,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 68335
+          "row_group": 0,
+          "row": 19577
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 6829828
+      "offset": 0,
+      "documents": 6871608
     },
     "cortex": {
-      "page": 214689,
-      "row": 86,
+      "page": 215998,
+      "row": 28,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 27480278
+      "documents": 27647772
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 820000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "be65b8dd85d32e3c046a0a29d18dd9c9636bd3518c6180704165ccc3b345f97a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "72a6fa391959ac41be57caac41e5d9e33fa242f668e7e959e3b49726861e8403", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "cca80237086820dcf5c494e00db56e92fb63f5bff088142a981f0d3db418bb87", "coverage.json": "c13407e1cced45349ea4cc9f52ed38e019bd1fa9d3e900afc3652ad06d436a2c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000820000"}
+{"step": 825000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "c94a248685b839a2edfdd0022e869649e1d13c7266eb40bee2cd5b65ffe131e7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "de221e9ed21e83c30b24525b864378be67505342cacb8dfb75a9ebfcd0672c2b", "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "3a5912414de65c80e1ba8c4b974f6204935ab0245d673f1014baff16d50c52ea", "coverage.json": "384700a0a13bd6ae2e77220c1d1c48f9e386f807b0b80377087bdf716560b06c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000825000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be65b8dd85d32e3c046a0a29d18dd9c9636bd3518c6180704165ccc3b345f97a
+oid sha256:c94a248685b839a2edfdd0022e869649e1d13c7266eb40bee2cd5b65ffe131e7
 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": 76942.25922787549,
-      "required_pages_per_second_upper_estimate": 0.005323044024022557
+      "required_token_positions_per_second": 75686.52414391923,
+      "required_pages_per_second_upper_estimate": 0.005236169357207637
     },
     {
       "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": 9617.782403484436,
-      "required_pages_per_second_upper_estimate": 0.004026764650203828
+      "required_token_positions_per_second": 9460.815517989904,
+      "required_pages_per_second_upper_estimate": 0.0039610458930885726
     },
     {
       "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": 19235.564806968872,
-      "required_pages_per_second_upper_estimate": 0.004006260873334826
+      "required_token_positions_per_second": 18921.631035979808,
+      "required_pages_per_second_upper_estimate": 0.003940876747828083
     },
     {
       "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": 28853.347210453307,
-      "required_pages_per_second_upper_estimate": 0.00172971410875644
+      "required_token_positions_per_second": 28382.44655396971,
+      "required_pages_per_second_upper_estimate": 0.0017014843334239179
     },
     {
       "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": 28853.347210453307,
-      "required_pages_per_second_upper_estimate": 0.0022432208331508883
+      "required_token_positions_per_second": 28382.44655396971,
+      "required_pages_per_second_upper_estimate": 0.0022066103783823774
     },
     {
       "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": 28853.347210453307,
-      "required_pages_per_second_upper_estimate": 0.37009975770517706
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28382.44655396971,
+      "required_pages_per_second_upper_estimate": 0.36405954969753734
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.08518336581272479,
-    "producer_p95_seconds": 0.2713440530023945,
-    "gpu_update_seconds": 0.958225047455963,
-    "producer_batches_per_second": 11.739381162731878,
-    "gpu_batches_per_second": 1.0435961809336411,
-    "producer_headroom": 11.248969071762392,
+    "producer_batch_seconds": 0.3011315108518602,
+    "producer_p95_seconds": 0.2949342350002553,
+    "gpu_update_seconds": 0.9741232119446381,
+    "producer_batches_per_second": 3.320808231497048,
+    "gpu_batches_per_second": 1.0265641838096684,
+    "producer_headroom": 3.2348763807180982,
     "sustainable": true,
-    "buffer_seconds": 60.36817798972567,
+    "buffer_seconds": 60.39563914056757,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 237498.7448770881,
-    "upload_bytes_per_second": 3973102.9977444257,
+    "generation_bytes_per_second": 533735.996489511,
+    "upload_bytes_per_second": 3295611.2481271825,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 50.96170196279986,
+    "estimated_full_shards_per_hour": 114.52732010854719,
     "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": 1.6904037299991614,
-      "bytes": 12097177,
+      "seconds": 1.3231591360017774,
+      "bytes": 1106930,
       "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 2.400186033999489,
+      "bytes": 7382284,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.2079874939990987,
-      "bytes": 1225243,
-      "pages": 35,
-      "shards": 1
+      "seconds": 2.1097241989991744,
+      "bytes": 16842562,
+      "pages": 44,
+      "shards": 5
     },
     {
-      "seconds": 1.7503577689985832,
-      "bytes": 11722749,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.6551350050031033,
+      "bytes": 2311821,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.717468271999678,
-      "bytes": 6201366,
+      "seconds": 1.769397926000238,
+      "bytes": 6044362,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4180366079999658,
-      "bytes": 818969,
-      "pages": 25,
-      "shards": 1
+      "seconds": 1.4953640580024512,
+      "bytes": 6379384,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.935665021999739,
-      "bytes": 7829235,
-      "pages": 39,
+      "seconds": 1.7474238050017448,
+      "bytes": 6290249,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.7852341269990575,
-      "bytes": 11660999,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.5282406320002337,
+      "bytes": 5939964,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.8856627909990493,
-      "bytes": 7118698,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.2798854280008527,
+      "bytes": 545338,
+      "pages": 18,
+      "shards": 1
+    },
+    {
+      "seconds": 1.8976086379989283,
+      "bytes": 13629725,
+      "pages": 49,
+      "shards": 4
+    },
+    {
+      "seconds": 1.6741066590002447,
+      "bytes": 5692006,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.6447518009990745,
-      "bytes": 6129393,
+      "seconds": 1.5403353149995382,
+      "bytes": 7551970,
       "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.5439765690025524,
+      "bytes": 6266973,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.743864690999544,
-      "bytes": 1054003,
-      "pages": 35,
+      "seconds": 1.459664652000356,
+      "bytes": 1024497,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.5393589430004795,
-      "bytes": 11424070,
-      "pages": 35,
+      "seconds": 1.8518995710001036,
+      "bytes": 10973445,
+      "pages": 22,
       "shards": 4
     },
     {
-      "seconds": 1.9194490070003667,
-      "bytes": 6288888,
-      "pages": 35,
+      "seconds": 1.523874344999058,
+      "bytes": 6793396,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.4686280049991183,
-      "bytes": 8470485,
+      "seconds": 1.551560215997597,
+      "bytes": 6347247,
       "pages": 37,
-      "shards": 4
+      "shards": 2
     },
     {
-      "seconds": 1.9930366190001223,
-      "bytes": 5866239,
-      "pages": 26,
+      "seconds": 1.3656726940025692,
+      "bytes": 2327821,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 2.011558195001271,
-      "bytes": 5842602,
-      "pages": 41,
-      "shards": 2
+      "seconds": 1.6191135740009486,
+      "bytes": 10591273,
+      "pages": 42,
+      "shards": 3
     },
     {
-      "seconds": 1.6693870859999151,
-      "bytes": 7612861,
+      "seconds": 1.5766091279983812,
+      "bytes": 7345474,
       "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.3870315900003334,
-      "bytes": 1261477,
+      "seconds": 1.195227366002655,
+      "bytes": 1189304,
       "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.6420892379992438,
-      "bytes": 7607765,
-      "pages": 39,
+      "seconds": 2.000728038001398,
+      "bytes": 11892571,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.5253431410001212,
-      "bytes": 5688723,
-      "pages": 19,
+      "seconds": 1.4507886779974797,
+      "bytes": 2119215,
+      "pages": 21,
       "shards": 2
     },
     {
-      "seconds": 1.593651556999248,
-      "bytes": 11614889,
-      "pages": 46,
+      "seconds": 1.5652879460030817,
+      "bytes": 11222900,
+      "pages": 47,
       "shards": 3
     },
     {
-      "seconds": 1.818591990000641,
-      "bytes": 7461958,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.6341935630007356,
+      "bytes": 5648901,
+      "pages": 19,
+      "shards": 2
     },
     {
-      "seconds": 1.4790617950002343,
-      "bytes": 1048131,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.5995988550021139,
+      "bytes": 2515627,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3689376410002296,
-      "bytes": 5834070,
-      "pages": 28,
+      "seconds": 1.3335332679998828,
+      "bytes": 5896707,
+      "pages": 29,
+      "shards": 2
+    },
+    {
+      "seconds": 1.9101363129993842,
+      "bytes": 4340262,
+      "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 1.2757619459989655,
-      "bytes": 12484592,
-      "pages": 45,
+      "seconds": 2.043565675998252,
+      "bytes": 13514454,
+      "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 1.4620103640008892,
-      "bytes": 316633,
-      "pages": 8,
-      "shards": 1
+      "seconds": 1.75038528100049,
+      "bytes": 5975710,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.8397933680007554,
-      "bytes": 6758434,
+      "seconds": 2.091957274999004,
+      "bytes": 2323345,
       "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.9822834630012949,
-      "bytes": 7671850,
-      "pages": 38,
+      "seconds": 1.9117556889978005,
+      "bytes": 10784078,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 1.43599132899908,
-      "bytes": 5917528,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.6485553290003736,
+      "bytes": 917525,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.7868269310019969,
-      "bytes": 5421842,
-      "pages": 40,
-      "shards": 2
+      "seconds": 2.0703852710030333,
+      "bytes": 13336105,
+      "pages": 42,
+      "shards": 4
     },
     {
-      "seconds": 1.8639257160029956,
-      "bytes": 7235620,
-      "pages": 29,
+      "seconds": 1.4275249089987483,
+      "bytes": 5696136,
+      "pages": 26,
+      "shards": 3
+    },
+    {
+      "seconds": 1.6552812369991443,
+      "bytes": 6589515,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.8414086459997634,
-      "bytes": 5860537,
+      "seconds": 1.7584444239982986,
+      "bytes": 6116166,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.8819817990006413,
-      "bytes": 6541422,
-      "pages": 23,
+      "seconds": 1.7439238019978802,
+      "bytes": 6171251,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.735375592001219,
-      "bytes": 12429197,
-      "pages": 48,
-      "shards": 4
+      "seconds": 1.5497890789993107,
+      "bytes": 2026869,
+      "pages": 23,
+      "shards": 2
     },
     {
-      "seconds": 1.311772810000548,
-      "bytes": 931562,
-      "pages": 26,
-      "shards": 1
+      "seconds": 2.0294503110017104,
+      "bytes": 16328104,
+      "pages": 41,
+      "shards": 4
     },
     {
-      "seconds": 1.873525888997392,
-      "bytes": 7538765,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.913415155999246,
+      "bytes": 2704493,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.4924008759990102,
-      "bytes": 6022523,
-      "pages": 32,
+      "seconds": 1.6119605509993562,
+      "bytes": 6041654,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.568882242001564,
-      "bytes": 7013425,
+      "seconds": 1.6828175019982154,
+      "bytes": 10664698,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.719156404997193,
-      "bytes": 11941861,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.565426788998593,
+      "bytes": 2415576,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.5756286089999776,
-      "bytes": 6163874,
+      "seconds": 1.5388405060002697,
+      "bytes": 6065532,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.85335265699905,
-      "bytes": 5996093,
+      "seconds": 1.3961634170009347,
+      "bytes": 973137,
       "pages": 28,
-      "shards": 3
-    },
-    {
-      "seconds": 1.6912986729985278,
-      "bytes": 6223259,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3231591360017774,
-      "bytes": 1106930,
-      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 2.400186033999489,
-      "bytes": 7382284,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 2.1097241989991744,
-      "bytes": 16842562,
+      "seconds": 2.0744263130000036,
+      "bytes": 17948904,
       "pages": 44,
       "shards": 5
     },
     {
-      "seconds": 1.6551350050031033,
-      "bytes": 2311821,
+      "seconds": 1.631489851999504,
+      "bytes": 5794503,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.769397926000238,
-      "bytes": 6044362,
+      "seconds": 1.5992371809988981,
+      "bytes": 2589704,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4953640580024512,
-      "bytes": 6379384,
-      "pages": 37,
+      "seconds": 1.4273169570005848,
+      "bytes": 6151606,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.7474238050017448,
-      "bytes": 6290249,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5282406320002337,
-      "bytes": 5939964,
-      "pages": 36,
+      "seconds": 1.4220550049976737,
+      "bytes": 4624242,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2798854280008527,
-      "bytes": 545338,
-      "pages": 18,
-      "shards": 1
+      "seconds": 2.047509969001112,
+      "bytes": 7597965,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.8976086379989283,
-      "bytes": 13629725,
-      "pages": 49,
-      "shards": 4
+      "seconds": 1.836660891000065,
+      "bytes": 12233872,
+      "pages": 41,
+      "shards": 3
     },
     {
-      "seconds": 1.6741066590002447,
-      "bytes": 5692006,
-      "pages": 27,
+      "seconds": 1.7772512390001793,
+      "bytes": 2617112,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.5403353149995382,
-      "bytes": 7551970,
-      "pages": 37,
+      "seconds": 1.5108418660020106,
+      "bytes": 11046966,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.5439765690025524,
-      "bytes": 6266973,
+      "seconds": 1.4716807509976206,
+      "bytes": 1094934,
       "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.459664652000356,
-      "bytes": 1024497,
-      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.8518995710001036,
-      "bytes": 10973445,
-      "pages": 22,
-      "shards": 4
-    },
-    {
-      "seconds": 1.523874344999058,
-      "bytes": 6793396,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.608142429999134,
+      "bytes": 7266564,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.551560215997597,
-      "bytes": 6347247,
-      "pages": 37,
+      "seconds": 1.734035388999473,
+      "bytes": 6027691,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3656726940025692,
-      "bytes": 2327821,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.5976459840021562,
+      "bytes": 11304870,
+      "pages": 44,
+      "shards": 3
     },
     {
-      "seconds": 1.6191135740009486,
-      "bytes": 10591273,
-      "pages": 42,
+      "seconds": 1.606354307001311,
+      "bytes": 7330991,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.5766091279983812,
-      "bytes": 7345474,
-      "pages": 29,
-      "shards": 3
+      "seconds": 1.5559647560003214,
+      "bytes": 6263907,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.195227366002655,
-      "bytes": 1189304,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.8874185050008236,
+      "bytes": 2469051,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 2.000728038001398,
-      "bytes": 11892571,
-      "pages": 34,
+      "seconds": 24.49332847400001,
+      "bytes": 9957694,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.4507886779974797,
-      "bytes": 2119215,
-      "pages": 21,
-      "shards": 2
+      "seconds": 1.9510848309982975,
+      "bytes": 11992256,
+      "pages": 40,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72a6fa391959ac41be57caac41e5d9e33fa242f668e7e959e3b49726861e8403
-size 721780159
+oid sha256:de221e9ed21e83c30b24525b864378be67505342cacb8dfb75a9ebfcd0672c2b
+size 721770943

training_config.json

@@ -53,7 +53,7 @@
   "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,