Qwen90M step 905000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 900000,
+  "step": 905000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7",
+    "model.safetensors": "2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d",
+    "trainer.pt": "3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4",
     "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9",
-    "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6",
+    "shard_scheduler.json": "b24e87ee5310f8cae440ce163c660971b4bd3a3c4202793a1fd52e7a8b911217",
+    "coverage.json": "2b0b0982d723d593e8b8cd110f0581bc41bb25cbc26ab9aa84796b740f5c7b28",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 900000,
+  "step": 905000,
   "stats": {
-    "climbmix/blocks": 32400000,
-    "climbmix/targets": 54162259750,
-    "climbmix/input_tokens": 54165692291,
-    "ultra_qa/blocks": 12150000,
-    "ultra_qa/targets": 19821847028,
-    "ultra_qa/input_tokens": 19821885096,
-    "cortex/blocks": 12150000,
-    "cortex/targets": 461992778,
-    "cortex/input_tokens": 19313376697,
-    "ultra_style/blocks": 12150000,
-    "ultra_style/targets": 19217238700,
-    "ultra_style/input_tokens": 19218246348,
-    "openmath/blocks": 8100000,
-    "openmath/targets": 12515849799,
-    "openmath/input_tokens": 12520019747,
-    "rewrite6/blocks": 4050000,
-    "rewrite6/targets": 6639430765,
-    "rewrite6/input_tokens": 6646710396
+    "climbmix/blocks": 32580000,
+    "climbmix/targets": 54463088890,
+    "climbmix/input_tokens": 54466540635,
+    "ultra_qa/blocks": 12217500,
+    "ultra_qa/targets": 19931980030,
+    "ultra_qa/input_tokens": 19932018303,
+    "cortex/blocks": 12217500,
+    "cortex/targets": 464554652,
+    "cortex/input_tokens": 19420649920,
+    "ultra_style/blocks": 12217500,
+    "ultra_style/targets": 19324167416,
+    "ultra_style/input_tokens": 19325180301,
+    "openmath/blocks": 8145000,
+    "openmath/targets": 12585378739,
+    "openmath/input_tokens": 12589571832,
+    "rewrite6/blocks": 4072500,
+    "rewrite6/targets": 6676302372,
+    "rewrite6/input_tokens": 6683622035
   },
   "sources": {
     "climbmix": {
-      "page": 4437,
-      "row": 1722,
+      "page": 4462,
+      "row": 661,
       "start": {
-        "file": 214,
+        "file": 215,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 65,
+          "row_group": 82,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 18175674
+      "offset": 0,
+      "documents": 18277013
     },
     "rewrite6": {
-      "page": 24,
-      "row": 2293,
+      "page": 38,
+      "row": 2160,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 98304
+          "row": 155648
         }
       },
       "epoch": 50,
-      "offset": 2048,
-      "documents": 10370797
+      "offset": 0,
+      "documents": 10428008
     },
     "openmath": {
-      "page": 2773,
-      "row": 839,
+      "page": 2788,
+      "row": 2605,
       "start": {
         "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 311
+          "row_group": 66,
+          "row": 751
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 11359047
+      "offset": 0,
+      "documents": 11422253
     },
     "ultra_qa": {
-      "page": 1199,
-      "row": 201,
+      "page": 1205,
+      "row": 2927,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 3,
-          "row": 40700
+          "row_group": 4,
+          "row": 744
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4911305
+      "documents": 4938607
     },
     "ultra_style": {
-      "page": 1830,
-      "row": 482,
+      "page": 1840,
+      "row": 498,
       "start": {
-        "file": 10,
+        "file": 11,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 50514
+          "row_group": 0,
+          "row": 2284
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 7496162
+      "offset": 0,
+      "documents": 7537138
     },
     "cortex": {
-      "page": 235633,
-      "row": 81,
+      "page": 236942,
+      "row": 28,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 30161105
+      "documents": 30328604
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 900000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9", "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000900000"}
+{"step": 905000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "b24e87ee5310f8cae440ce163c660971b4bd3a3c4202793a1fd52e7a8b911217", "coverage.json": "2b0b0982d723d593e8b8cd110f0581bc41bb25cbc26ab9aa84796b740f5c7b28", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000905000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7
+oid sha256:2eafced5169e47d39f63c15f6d5c04b0ee60f62d179c1e2dbe090427243f8630
 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": 77365.33532253442,
-      "required_pages_per_second_upper_estimate": 0.005352313409922849
+      "required_token_positions_per_second": 76565.79414645793,
+      "required_pages_per_second_upper_estimate": 0.005296999296170754
     },
     {
       "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": 9670.666915316802,
-      "required_pages_per_second_upper_estimate": 0.004048906290953853
+      "required_token_positions_per_second": 9570.724268307242,
+      "required_pages_per_second_upper_estimate": 0.0040070623916907445
     },
     {
       "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": 19341.333830633605,
-      "required_pages_per_second_upper_estimate": 0.004028289771647462
+      "required_token_positions_per_second": 19141.448536614484,
+      "required_pages_per_second_upper_estimate": 0.003986658936232964
     },
     {
       "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": 29012.00074595041,
-      "required_pages_per_second_upper_estimate": 0.0017392251459595694
+      "required_token_positions_per_second": 28712.172804921724,
+      "required_pages_per_second_upper_estimate": 0.0017212509187056575
     },
     {
       "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": 29012.00074595041,
-      "required_pages_per_second_upper_estimate": 0.002255555447692636
+      "required_token_positions_per_second": 28712.172804921724,
+      "required_pages_per_second_upper_estimate": 0.0022322451440813925
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "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": 29012.00074595041,
-      "required_pages_per_second_upper_estimate": 0.3721347949096395
+      "required_token_positions_per_second": 28712.172804921724,
+      "required_pages_per_second_upper_estimate": 0.36828892401228464
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.10793024863289702,
-    "producer_p95_seconds": 0.2442270199990162,
-    "gpu_update_seconds": 0.9529849472328861,
-    "producer_batches_per_second": 9.265243179428767,
-    "gpu_batches_per_second": 1.0493345177210072,
-    "producer_headroom": 8.82963728244778,
+    "producer_batch_seconds": 0.067691734976421,
+    "producer_p95_seconds": 0.24887008000223432,
+    "gpu_update_seconds": 0.9629365282748887,
+    "producer_batches_per_second": 14.772852259560626,
+    "gpu_batches_per_second": 1.03849004647431,
+    "producer_headroom": 14.225319067539154,
     "sustainable": true,
-    "buffer_seconds": 60.038051675671824,
+    "buffer_seconds": 60.66500128131799,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 83,
+    "upload_pages_per_commit": 86,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 398086.69224948075,
-    "upload_bytes_per_second": 5232468.799184691,
+    "generation_bytes_per_second": 346296.08435846475,
+    "upload_bytes_per_second": 5025204.241395261,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 85.42013717282597,
+    "estimated_full_shards_per_hour": 74.30707834306199,
     "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": 1.1932894399997167,
-      "bytes": 6531618,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.3964128229999915,
+      "bytes": 12688722,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.1864704220006388,
-      "bytes": 6136727,
-      "pages": 35,
+      "seconds": 1.4358708510008,
+      "bytes": 6164418,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2697396670000671,
-      "bytes": 6182466,
-      "pages": 33,
+      "seconds": 1.5897931969993806,
+      "bytes": 7707081,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.2363136280000617,
-      "bytes": 6022734,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.0018968719996337,
+      "bytes": 927002,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.2937560119999034,
-      "bytes": 947666,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.3852189129993349,
+      "bytes": 10288556,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.5761228700002903,
-      "bytes": 18281686,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.2743636189989047,
+      "bytes": 2631065,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.1403170100002171,
-      "bytes": 5965661,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.369053440999778,
+      "bytes": 11976832,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.088612870000361,
-      "bytes": 2479743,
-      "pages": 29,
+      "seconds": 1.3680221860013262,
+      "bytes": 5930978,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.3296366939994186,
-      "bytes": 6147791,
+      "seconds": 1.3103218420001213,
+      "bytes": 6746870,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4409255560003658,
+      "bytes": 6062763,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1267749469989212,
-      "bytes": 894977,
-      "pages": 27,
+      "seconds": 1.1668520560015168,
+      "bytes": 7696896,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1897505889992317,
+      "bytes": 1090330,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.4399002789996302,
-      "bytes": 10842331,
-      "pages": 40,
+      "seconds": 1.3477415189991007,
+      "bytes": 15713674,
+      "pages": 35,
       "shards": 4
     },
     {
-      "seconds": 1.632049243999063,
-      "bytes": 11967346,
-      "pages": 40,
+      "seconds": 1.2074203040010616,
+      "bytes": 7463350,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9512228559997311,
-      "bytes": 901267,
-      "pages": 26,
+      "seconds": 1.190478869000799,
+      "bytes": 1226986,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.5480925499996374,
+      "bytes": 11721285,
+      "pages": 32,
+      "shards": 4
+    },
+    {
+      "seconds": 1.0147747470000468,
+      "bytes": 1162064,
+      "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.6983987419989717,
-      "bytes": 12559131,
+      "seconds": 1.132629542000359,
+      "bytes": 6067377,
+      "pages": 33,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3739411900005507,
+      "bytes": 13347857,
       "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 1.1707839329992566,
-      "bytes": 6060902,
+      "seconds": 1.5033581300012884,
+      "bytes": 979335,
       "pages": 32,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.1801391750013863,
-      "bytes": 2749619,
-      "pages": 37,
-      "shards": 3
+      "seconds": 1.363758608000353,
+      "bytes": 11974676,
+      "pages": 44,
+      "shards": 4
     },
     {
-      "seconds": 1.1686728980002954,
-      "bytes": 6121581,
+      "seconds": 1.1600040840003203,
+      "bytes": 6119662,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3895534080002108,
-      "bytes": 15955272,
-      "pages": 35,
+      "seconds": 0.8502458040002239,
+      "bytes": 891680,
+      "pages": 27,
+      "shards": 1
+    },
+    {
+      "seconds": 1.5061208509996504,
+      "bytes": 11821724,
+      "pages": 33,
       "shards": 4
     },
     {
-      "seconds": 0.9986490080009389,
-      "bytes": 2653371,
-      "pages": 37,
+      "seconds": 1.1246318519988563,
+      "bytes": 6660673,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.1648389490001136,
-      "bytes": 5951393,
+      "seconds": 1.2537715590005973,
+      "bytes": 5914249,
       "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.1927761070000997,
-      "bytes": 2669572,
+      "seconds": 1.3754841779991693,
+      "bytes": 2754086,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.24713432200042,
+      "bytes": 5975341,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1629383399995277,
-      "bytes": 10339749,
-      "pages": 40,
+      "seconds": 1.473726465999789,
+      "bytes": 9735362,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.6129309269999794,
-      "bytes": 5876330,
-      "pages": 28,
+      "seconds": 1.1486626710011478,
+      "bytes": 2585051,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.7068213439997635,
-      "bytes": 8291481,
-      "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2685745549988496,
-      "bytes": 6220542,
+      "seconds": 1.5323144779995346,
+      "bytes": 11971610,
       "pages": 34,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.144102094000118,
-      "bytes": 5891615,
-      "pages": 30,
+      "seconds": 1.079354624998814,
+      "bytes": 2321724,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.2515704560009908,
-      "bytes": 7512167,
-      "pages": 38,
+      "seconds": 1.6746519440002885,
+      "bytes": 10811116,
+      "pages": 42,
       "shards": 3
     },
     {
-      "seconds": 1.4014940459983336,
-      "bytes": 5931776,
-      "pages": 29,
+      "seconds": 1.732792734999748,
+      "bytes": 6311900,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1321277869992628,
-      "bytes": 2736110,
-      "pages": 36,
+      "seconds": 1.1773680320002313,
+      "bytes": 2462043,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.411381372001415,
-      "bytes": 15991084,
+      "seconds": 1.2794374899985996,
+      "bytes": 6045695,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "seconds": 1.710491572997853,
+      "bytes": 17528313,
       "pages": 40,
-      "shards": 4
+      "shards": 5
     },
     {
-      "seconds": 0.8898140989986132,
-      "bytes": 1156490,
-      "pages": 32,
+      "seconds": 1.008943750999606,
+      "bytes": 1209380,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.3950513529998716,
-      "bytes": 7312918,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.5094498580001527,
+      "bytes": 5827665,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.2647196729994903,
-      "bytes": 6433157,
-      "pages": 36,
+      "seconds": 1.0169636540013016,
+      "bytes": 2368292,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2063595479994547,
-      "bytes": 6907543,
-      "pages": 33,
+      "seconds": 1.3777095789992018,
+      "bytes": 11141641,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.290804393000144,
-      "bytes": 6092692,
+      "seconds": 3.434384232998127,
+      "bytes": 5971898,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.4008768420007982,
-      "bytes": 11876531,
-      "pages": 36,
+      "seconds": 1.1304158860002644,
+      "bytes": 8230136,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9798814419991686,
-      "bytes": 2581068,
-      "pages": 36,
+      "seconds": 1.2011196589992323,
+      "bytes": 6075797,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.9553985610000382,
-      "bytes": 10094743,
-      "pages": 34,
+      "seconds": 1.276424440999108,
+      "bytes": 6070440,
+      "pages": 25,
       "shards": 3
     },
     {
-      "seconds": 1.0605444640004862,
-      "bytes": 2480385,
-      "pages": 33,
+      "seconds": 1.2386367050021363,
+      "bytes": 6009209,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.4249819679989741,
-      "bytes": 6270908,
-      "pages": 32,
+      "seconds": 1.1234037470021576,
+      "bytes": 6172938,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2829982630009908,
-      "bytes": 5992379,
-      "pages": 37,
+      "seconds": 0.8850959770024929,
+      "bytes": 2438669,
+      "pages": 31,
       "shards": 2
     },
     {
-      "seconds": 1.3964128229999915,
-      "bytes": 12688722,
-      "pages": 35,
+      "seconds": 1.6140179690009973,
+      "bytes": 15868305,
+      "pages": 31,
       "shards": 4
     },
     {
-      "seconds": 1.4358708510008,
-      "bytes": 6164418,
-      "pages": 37,
+      "seconds": 1.348259312002483,
+      "bytes": 5928296,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.5897931969993806,
-      "bytes": 7707081,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0018968719996337,
-      "bytes": 927002,
-      "pages": 29,
-      "shards": 1
+      "seconds": 0.9825653119987692,
+      "bytes": 2330980,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.3852189129993349,
-      "bytes": 10288556,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.2888210940000135,
+      "bytes": 6181272,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.2743636189989047,
-      "bytes": 2631065,
-      "pages": 37,
+      "seconds": 1.1400288160002674,
+      "bytes": 6499509,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.369053440999778,
-      "bytes": 11976832,
+      "seconds": 1.1078390350012342,
+      "bytes": 6180333,
       "pages": 37,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.3680221860013262,
-      "bytes": 5930978,
-      "pages": 33,
+      "seconds": 1.1313278150009864,
+      "bytes": 6410101,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.3103218420001213,
-      "bytes": 6746870,
-      "pages": 31,
+      "seconds": 1.2548322069997084,
+      "bytes": 7827361,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.4409255560003658,
-      "bytes": 6062763,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1668520560015168,
-      "bytes": 7696896,
-      "pages": 38,
+      "seconds": 1.3341312570009904,
+      "bytes": 10704029,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.1897505889992317,
-      "bytes": 1090330,
-      "pages": 30,
+      "seconds": 1.136130429997138,
+      "bytes": 965107,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.3477415189991007,
-      "bytes": 15713674,
-      "pages": 35,
-      "shards": 4
-    },
-    {
-      "seconds": 1.2074203040010616,
-      "bytes": 7463350,
-      "pages": 39,
+      "seconds": 1.3451466709993838,
+      "bytes": 7707115,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.190478869000799,
-      "bytes": 1226986,
-      "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.5480925499996374,
-      "bytes": 11721285,
-      "pages": 32,
-      "shards": 4
+      "seconds": 2.3437790619973384,
+      "bytes": 11207805,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.0147747470000468,
-      "bytes": 1162064,
-      "pages": 33,
-      "shards": 1
+      "seconds": 2.2792270009995264,
+      "bytes": 8233418,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.132629542000359,
-      "bytes": 6067377,
-      "pages": 33,
+      "seconds": 1.1861695589977899,
+      "bytes": 6023568,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3739411900005507,
-      "bytes": 13347857,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 1.5033581300012884,
-      "bytes": 979335,
-      "pages": 32,
+      "seconds": 1.0451825370000734,
+      "bytes": 965218,
+      "pages": 28,
       "shards": 1
     },
     {
-      "seconds": 1.363758608000353,
-      "bytes": 11974676,
-      "pages": 44,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1600040840003203,
-      "bytes": 6119662,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8502458040002239,
-      "bytes": 891680,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.2396269109995046,
+      "bytes": 7721145,
+      "pages": 37,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d
-size 721766271
+oid sha256:3c5628671e1b2c65e85ea6083662d90b0c22e2a2bf9df3cd5c38a446fd7e4fa4
+size 721773631