Qwen90M step 580000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 575000,
+  "step": 580000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d",
+    "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa",
+    "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964",
     "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
     "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
-    "shard_scheduler.json": "96d62a23720c59296a4eede4e35a800f12a60c2214c596aa38c0fd0f3fb5fd82",
-    "coverage.json": "a6dd49afdfac1374a27b1c3438e7f2ed72df001cb3fb4b33fb5a8cf83ee869d9",
+    "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb",
+    "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 575000,
+  "step": 580000,
   "stats": {
-    "climbmix/blocks": 20700000,
-    "climbmix/targets": 34602598775,
-    "climbmix/input_tokens": 34604791626,
-    "ultra_qa/blocks": 7762500,
-    "ultra_qa/targets": 12663846760,
-    "ultra_qa/input_tokens": 12663871196,
-    "cortex/blocks": 7762500,
-    "cortex/targets": 295179734,
-    "cortex/input_tokens": 12339109967,
-    "ultra_style/blocks": 7762500,
-    "ultra_style/targets": 12277961225,
-    "ultra_style/input_tokens": 12278605070,
-    "openmath/blocks": 5175000,
-    "openmath/targets": 7996288244,
-    "openmath/input_tokens": 7998951398,
-    "rewrite6/blocks": 2587500,
-    "rewrite6/targets": 4241858572,
-    "rewrite6/input_tokens": 4246508964
+    "climbmix/blocks": 20880000,
+    "climbmix/targets": 34903766814,
+    "climbmix/input_tokens": 34905978789,
+    "ultra_qa/blocks": 7830000,
+    "ultra_qa/targets": 12773998197,
+    "ultra_qa/input_tokens": 12774022841,
+    "cortex/blocks": 7830000,
+    "cortex/targets": 297744461,
+    "cortex/input_tokens": 12446411636,
+    "ultra_style/blocks": 7830000,
+    "ultra_style/targets": 12384628359,
+    "ultra_style/input_tokens": 12385277999,
+    "openmath/blocks": 5220000,
+    "openmath/targets": 8065835620,
+    "openmath/input_tokens": 8068521894,
+    "rewrite6/blocks": 2610000,
+    "rewrite6/targets": 4278710471,
+    "rewrite6/input_tokens": 4283401756
   },
   "sources": {
     "climbmix": {
-      "page": 2835,
-      "row": 588,
+      "page": 2859,
+      "row": 2966,
       "start": {
-        "file": 137,
+        "file": 138,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 16,
+          "row_group": 29,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 11612748
+      "offset": 10240,
+      "documents": 11713430
     },
     "rewrite6": {
-      "page": 12,
-      "row": 3116,
+      "page": 26,
+      "row": 3974,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 49152
+          "row": 106496
         }
       },
       "epoch": 32,
-      "offset": 2048,
-      "documents": 6625196
+      "offset": 0,
+      "documents": 6683398
     },
     "openmath": {
-      "page": 1771,
-      "row": 1892,
+      "page": 1786,
+      "row": 3533,
       "start": {
         "file": 16,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 267,
-          "row": 616
+          "row_group": 329,
+          "row": 56
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 7255908
+      "offset": 2048,
+      "documents": 7318989
     },
     "ultra_qa": {
-      "page": 766,
-      "row": 264,
+      "page": 772,
+      "row": 2970,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 19779
+          "row": 44355
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 3137800
+      "documents": 3165082
     },
     "ultra_style": {
-      "page": 1169,
-      "row": 508,
+      "page": 1179,
+      "row": 1648,
       "start": {
-        "file": 6,
+        "file": 7,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 82661
+          "row_group": 0,
+          "row": 34590
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 4788732
+      "offset": 2048,
+      "documents": 4830832
     },
     "cortex": {
-      "page": 150548,
-      "row": 11,
+      "page": 151857,
+      "row": 6,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 19270155
+      "documents": 19437702
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 575000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "96d62a23720c59296a4eede4e35a800f12a60c2214c596aa38c0fd0f3fb5fd82", "coverage.json": "a6dd49afdfac1374a27b1c3438e7f2ed72df001cb3fb4b33fb5a8cf83ee869d9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000575000"}
+{"step": 580000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb", "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000580000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d
+oid sha256:86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2
 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": 76608.2440593413,
-      "required_pages_per_second_upper_estimate": 0.005299936079641394
+      "required_token_positions_per_second": 76844.82269293365,
+      "required_pages_per_second_upper_estimate": 0.0053163031384513715
     },
     {
       "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": 9576.030507417663,
-      "required_pages_per_second_upper_estimate": 0.004009284003199404
+      "required_token_positions_per_second": 9605.602836616707,
+      "required_pages_per_second_upper_estimate": 0.004021665319894588
     },
     {
       "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": 19152.061014835326,
-      "required_pages_per_second_upper_estimate": 0.003988869235576494
+      "required_token_positions_per_second": 19211.205673233413,
+      "required_pages_per_second_upper_estimate": 0.004001187508171285
     },
     {
       "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": 28728.09152225299,
-      "required_pages_per_second_upper_estimate": 0.001722205221503193
+      "required_token_positions_per_second": 28816.808509850118,
+      "required_pages_per_second_upper_estimate": 0.0017275236694466443
     },
     {
       "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": 28728.09152225299,
-      "required_pages_per_second_upper_estimate": 0.002233482754334861
+      "required_token_positions_per_second": 28816.808509850118,
+      "required_pages_per_second_upper_estimate": 0.0022403801098957493
     },
     {
       "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": 28728.09152225299,
-      "required_pages_per_second_upper_estimate": 0.36849311222602316
+      "required_token_positions_per_second": 28816.808509850118,
+      "required_pages_per_second_upper_estimate": 0.3696310784860394
     }
   ],
   "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": 704566,
-    "producer_batch_seconds": 0.6385047882343144,
-    "producer_p95_seconds": 0.21890624099978595,
-    "gpu_update_seconds": 0.962402948994494,
-    "producer_batches_per_second": 1.5661589676803274,
-    "gpu_batches_per_second": 1.0390658102666737,
-    "producer_headroom": 1.5072760090897195,
+    "producer_batch_seconds": 0.0661616753671126,
+    "producer_p95_seconds": 0.24438537600144628,
+    "gpu_update_seconds": 0.9594400431452845,
+    "producer_batches_per_second": 15.11449029141539,
+    "gpu_batches_per_second": 1.0422746133481668,
+    "producer_headroom": 14.501447217314565,
     "sustainable": true,
-    "buffer_seconds": 60.63138578665312,
+    "buffer_seconds": 60.444722718152924,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 310794.11682017095,
-    "upload_bytes_per_second": 5262266.363159667,
+    "generation_bytes_per_second": 406720.83803339687,
+    "upload_bytes_per_second": 5366265.038987921,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 66.68918255285116,
+    "estimated_full_shards_per_hour": 87.2728238654273,
     "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": 1.4154219219999504,
-      "bytes": 7407761,
-      "pages": 38,
+      "seconds": 1.2003755249988899,
+      "bytes": 10273667,
+      "pages": 44,
       "shards": 3
     },
     {
-      "seconds": 1.1865251550007088,
-      "bytes": 5346992,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0492023780007003,
-      "bytes": 6288876,
-      "pages": 37,
+      "seconds": 1.0975800400010485,
+      "bytes": 2683103,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.7196641009995801,
-      "bytes": 2291853,
+      "seconds": 1.1853528239989828,
+      "bytes": 5804023,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0487870670003758,
-      "bytes": 11950097,
-      "pages": 41,
+      "seconds": 1.138313076000486,
+      "bytes": 10599364,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.2208401020016026,
-      "bytes": 11777262,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9635194839993346,
-      "bytes": 1029009,
-      "pages": 33,
-      "shards": 1
-    },
-    {
-      "seconds": 1.0835146409990557,
-      "bytes": 5946250,
-      "pages": 30,
+      "seconds": 0.9668768979991,
+      "bytes": 2353349,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 0.9391229799985013,
-      "bytes": 2632765,
-      "pages": 34,
-      "shards": 2
+      "seconds": 1.4940221720007685,
+      "bytes": 12237361,
+      "pages": 41,
+      "shards": 3
     },
     {
-      "seconds": 1.0898778940008924,
-      "bytes": 10697456,
-      "pages": 35,
+      "seconds": 0.8858112369998707,
+      "bytes": 2416252,
+      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 1.3169437420001486,
-      "bytes": 11722554,
-      "pages": 35,
+      "seconds": 1.2992984630000137,
+      "bytes": 10017513,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.2027081380001619,
-      "bytes": 2601682,
-      "pages": 37,
+      "seconds": 4.1799396490005165,
+      "bytes": 5932760,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.0705139929996221,
-      "bytes": 6406862,
-      "pages": 35,
+      "seconds": 0.8289485320001404,
+      "bytes": 2513176,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 3.478892785999051,
-      "bytes": 11471341,
-      "pages": 38,
-      "shards": 4
+      "seconds": 0.9106573579992983,
+      "bytes": 5965787,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.575513838999541,
+      "bytes": 18103820,
+      "pages": 39,
+      "shards": 5
     },
     {
-      "seconds": 0.9223496029990201,
-      "bytes": 897417,
+      "seconds": 0.8145781550010724,
+      "bytes": 930729,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.0840730929994606,
-      "bytes": 6199927,
-      "pages": 34,
+      "seconds": 1.3511633730013273,
+      "bytes": 6255780,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.112097929999436,
-      "bytes": 8142857,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.031140414001129,
+      "bytes": 2762787,
+      "pages": 38,
+      "shards": 2
     },
     {
-      "seconds": 1.17854872099997,
-      "bytes": 10442242,
-      "pages": 43,
-      "shards": 4
+      "seconds": 1.0752210870014096,
+      "bytes": 5610482,
+      "pages": 19,
+      "shards": 2
     },
     {
-      "seconds": 1.163782216999607,
-      "bytes": 7671664,
-      "pages": 39,
+      "seconds": 1.425927022999531,
+      "bytes": 10079783,
+      "pages": 46,
       "shards": 3
     },
     {
-      "seconds": 1.1290235629985546,
-      "bytes": 1186173,
+      "seconds": 1.1401308700005757,
+      "bytes": 8309975,
       "pages": 34,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2037071409995406,
-      "bytes": 10277758,
-      "pages": 28,
       "shards": 3
     },
     {
-      "seconds": 0.9222849990001123,
-      "bytes": 2428641,
-      "pages": 31,
+      "seconds": 1.2373918849989423,
+      "bytes": 6097119,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1589227389995358,
-      "bytes": 11976819,
-      "pages": 37,
+      "seconds": 1.1592294109977956,
+      "bytes": 6491968,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.0021623419997923,
-      "bytes": 6273488,
-      "pages": 37,
+      "seconds": 1.6240387940015353,
+      "bytes": 5730323,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0131233649990463,
-      "bytes": 2204351,
-      "pages": 24,
+      "seconds": 1.3926623539991851,
+      "bytes": 6408754,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1051343430008274,
-      "bytes": 10092152,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.260699045000365,
+      "bytes": 2670941,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 0.975858148000043,
-      "bytes": 861206,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.4827766799971869,
+      "bytes": 15946512,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.055939247000424,
-      "bytes": 7659150,
+      "seconds": 1.33145204000175,
+      "bytes": 7515329,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1426443119999021,
-      "bytes": 11681956,
+      "seconds": 0.9712299829989206,
+      "bytes": 867774,
       "pages": 27,
-      "shards": 3
+      "shards": 1
     },
     {
-      "seconds": 2.1566952029988897,
-      "bytes": 6925851,
-      "pages": 39,
+      "seconds": 1.1689961140000378,
+      "bytes": 10341916,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9256938060007087,
-      "bytes": 6110229,
-      "pages": 34,
+      "seconds": 4.051943458998721,
+      "bytes": 2586599,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.7895571030003339,
-      "bytes": 1089032,
-      "pages": 34,
+      "seconds": 0.9690040009991208,
+      "bytes": 5902816,
+      "pages": 29,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9907326490028936,
+      "bytes": 186236,
+      "pages": 7,
       "shards": 1
     },
     {
-      "seconds": 1.843757688999176,
-      "bytes": 12452068,
-      "pages": 39,
-      "shards": 4
+      "seconds": 1.1790409999994154,
+      "bytes": 12018438,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 2.206128002000696,
-      "bytes": 6045751,
-      "pages": 29,
+      "seconds": 1.0833680409996305,
+      "bytes": 2366981,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.4661580640004104,
-      "bytes": 8583866,
-      "pages": 41,
+      "seconds": 1.2835563219996402,
+      "bytes": 10401632,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0245029070010787,
-      "bytes": 5870358,
-      "pages": 27,
-      "shards": 2
+      "seconds": 0.9978380930006097,
+      "bytes": 912994,
+      "pages": 28,
+      "shards": 1
     },
     {
-      "seconds": 0.900488690000202,
-      "bytes": 1094466,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1661826020026638,
+      "bytes": 7439707,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.2211129879997316,
-      "bytes": 12049529,
-      "pages": 44,
-      "shards": 4
+      "seconds": 1.0849376400001347,
+      "bytes": 10733196,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.1176546409988077,
-      "bytes": 5812216,
-      "pages": 27,
+      "seconds": 1.2397475959987787,
+      "bytes": 8159623,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2689961300020514,
+      "bytes": 6434173,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.8849807390015485,
-      "bytes": 1252571,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3497664780006744,
+      "bytes": 7459096,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.2397414250008296,
-      "bytes": 13176621,
+      "seconds": 1.0805805290001445,
+      "bytes": 1076352,
       "pages": 29,
-      "shards": 4
+      "shards": 1
     },
     {
-      "seconds": 1.2003755249988899,
-      "bytes": 10273667,
-      "pages": 44,
+      "seconds": 1.1792509940023592,
+      "bytes": 9839539,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.0975800400010485,
-      "bytes": 2683103,
-      "pages": 36,
+      "seconds": 1.0725541200008593,
+      "bytes": 2601868,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.1853528239989828,
-      "bytes": 5804023,
+      "seconds": 1.330655336001655,
+      "bytes": 11531295,
       "pages": 28,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.138313076000486,
-      "bytes": 10599364,
-      "pages": 39,
+      "seconds": 1.5732356919979793,
+      "bytes": 10605987,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 0.9668768979991,
-      "bytes": 2353349,
-      "pages": 29,
+      "seconds": 0.9885955910031043,
+      "bytes": 2380823,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.4940221720007685,
-      "bytes": 12237361,
-      "pages": 41,
-      "shards": 3
+      "seconds": 0.9730731119998381,
+      "bytes": 6090581,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 0.8858112369998707,
-      "bytes": 2416252,
-      "pages": 28,
+      "seconds": 1.1632558510027593,
+      "bytes": 7851564,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2992984630000137,
-      "bytes": 10017513,
-      "pages": 40,
-      "shards": 3
+      "seconds": 1.0146159530013392,
+      "bytes": 5636542,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 4.1799396490005165,
-      "bytes": 5932760,
-      "pages": 32,
-      "shards": 2
+      "seconds": 1.0724632399978873,
+      "bytes": 12061908,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 0.8289485320001404,
-      "bytes": 2513176,
-      "pages": 33,
+      "seconds": 0.9588815639981476,
+      "bytes": 2343317,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.9106573579992983,
-      "bytes": 5965787,
+      "seconds": 1.1246145429977332,
+      "bytes": 6175304,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.575513838999541,
-      "bytes": 18103820,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.2072375589996227,
+      "bytes": 11138063,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 0.8145781550010724,
-      "bytes": 930729,
-      "pages": 26,
+      "seconds": 0.9064097570008016,
+      "bytes": 1171302,
+      "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.3511633730013273,
-      "bytes": 6255780,
+      "seconds": 0.9584431869989203,
+      "bytes": 6298065,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.031140414001129,
-      "bytes": 2762787,
-      "pages": 38,
-      "shards": 2
+      "seconds": 0.972795835998113,
+      "bytes": 8254778,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.0752210870014096,
-      "bytes": 5610482,
-      "pages": 19,
-      "shards": 2
+      "seconds": 1.3151332330016885,
+      "bytes": 10337557,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.425927022999531,
-      "bytes": 10079783,
-      "pages": 46,
+      "seconds": 1.029109121998772,
+      "bytes": 7643162,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1401308700005757,
-      "bytes": 8309975,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.8460896450014843,
+      "bytes": 884554,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.2373918849989423,
-      "bytes": 6097119,
-      "pages": 33,
+      "seconds": 1.0631770070031052,
+      "bytes": 6064387,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.1592294109977956,
-      "bytes": 6491968,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.2800072050013114,
+      "bytes": 12000660,
+      "pages": 43,
+      "shards": 4
     },
     {
-      "seconds": 1.6240387940015353,
-      "bytes": 5730323,
-      "pages": 27,
+      "seconds": 1.2026952199994412,
+      "bytes": 6686271,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3926623539991851,
-      "bytes": 6408754,
+      "seconds": 1.3011887590000697,
+      "bytes": 6177443,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.260699045000365,
-      "bytes": 2670941,
-      "pages": 37,
+      "seconds": 1.0062806160021864,
+      "bytes": 2394266,
+      "pages": 29,
       "shards": 2
+    },
+    {
+      "seconds": 1.1657910530011577,
+      "bytes": 10819661,
+      "pages": 44,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa
-size 721756799
+oid sha256:1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964
+size 721773055