Qwen90M step 930000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 925000,
+  "step": 930000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2",
+    "model.safetensors": "63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520",
+    "trainer.pt": "c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56",
     "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
-    "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9",
-    "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5",
+    "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4",
+    "shard_scheduler.json": "b3c49f97562f447c57be5ea107f2e5123930591bd4d8ad5d22e1610da58f2a24",
+    "coverage.json": "32cef4aca490da53b293a3ee6d9244fadb87af3f5c0e004c874df34f97a8fb94",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 925000,
+  "step": 930000,
   "stats": {
-    "climbmix/blocks": 33300000,
-    "climbmix/targets": 55667089526,
-    "climbmix/input_tokens": 55670617480,
-    "ultra_qa/blocks": 12487500,
-    "ultra_qa/targets": 20372477924,
-    "ultra_qa/input_tokens": 20372517045,
-    "cortex/blocks": 12487500,
-    "cortex/targets": 474833481,
-    "cortex/input_tokens": 19849928864,
-    "ultra_style/blocks": 12487500,
-    "ultra_style/targets": 19750266654,
-    "ultra_style/input_tokens": 19751301641,
-    "openmath/blocks": 8325000,
-    "openmath/targets": 12863641674,
-    "openmath/input_tokens": 12867927568,
-    "rewrite6/blocks": 4162500,
-    "rewrite6/targets": 6823898189,
-    "rewrite6/input_tokens": 6831379657
+    "climbmix/blocks": 33480000,
+    "climbmix/targets": 55968202162,
+    "climbmix/input_tokens": 55971749108,
+    "ultra_qa/blocks": 12555000,
+    "ultra_qa/targets": 20482474984,
+    "ultra_qa/input_tokens": 20482514313,
+    "cortex/blocks": 12555000,
+    "cortex/targets": 477403369,
+    "cortex/input_tokens": 19957266714,
+    "ultra_style/blocks": 12555000,
+    "ultra_style/targets": 19857263635,
+    "ultra_style/input_tokens": 19858304258,
+    "openmath/blocks": 8370000,
+    "openmath/targets": 12933151733,
+    "openmath/input_tokens": 12937461063,
+    "rewrite6/blocks": 4185000,
+    "rewrite6/targets": 6860771033,
+    "rewrite6/input_tokens": 6868292697
   },
   "sources": {
     "climbmix": {
-      "page": 4560,
-      "row": 2473,
+      "page": 4585,
+      "row": 1014,
       "start": {
-        "file": 220,
+        "file": 221,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 62,
+          "row_group": 79,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 18680233
+      "documents": 18781174
     },
     "rewrite6": {
-      "page": 44,
-      "row": 2622,
+      "page": 8,
+      "row": 1951,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 180224
+          "row": 32768
         }
       },
-      "epoch": 51,
-      "offset": 4096,
-      "documents": 10658450
+      "epoch": 52,
+      "offset": 0,
+      "documents": 10715727
     },
     "openmath": {
-      "page": 2850,
-      "row": 1200,
+      "page": 2865,
+      "row": 3170,
       "start": {
         "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 320,
-          "row": 703
+          "row_group": 382,
+          "row": 143
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11674800
+      "documents": 11738210
     },
     "ultra_qa": {
-      "page": 1232,
-      "row": 1402,
+      "page": 1239,
+      "row": 96,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 46700
+          "row_group": 6,
+          "row": 10847
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 5047674
+      "documents": 5075040
     },
     "ultra_style": {
-      "page": 1880,
-      "row": 3523,
+      "page": 1890,
+      "row": 4094,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 81311
+          "row_group": 2,
+          "row": 36445
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 7704003
+      "documents": 7745534
     },
     "cortex": {
-      "page": 242179,
-      "row": 111,
+      "page": 243489,
+      "row": 37,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 30999023
+      "documents": 31166629
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 925000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9", "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000925000"}
+{"step": 930000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "b3c49f97562f447c57be5ea107f2e5123930591bd4d8ad5d22e1610da58f2a24", "coverage.json": "32cef4aca490da53b293a3ee6d9244fadb87af3f5c0e004c874df34f97a8fb94", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000930000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2
+oid sha256:63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414
 size 360622744

performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 90,
     "mode": "compiled",
-    "microbatch_seconds": 0.9382242159999805,
+    "microbatch_seconds": 0.9440444560000287,
     "peak_gb": 89.37790203094482,
     "fits_budget": true,
-    "estimated_update_seconds": 0.9382242159999805,
-    "input_tokens_per_second": 196456.2381323185
+    "estimated_update_seconds": 0.9440444560000287,
+    "input_tokens_per_second": 195245.04257031978
   },
   "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": 76780.75659017764,
-      "required_pages_per_second_upper_estimate": 0.005311870896808872
+      "required_token_positions_per_second": 77309.27550044844,
+      "required_pages_per_second_upper_estimate": 0.005348435061354233
     },
     {
       "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": 9597.594573772205,
-      "required_pages_per_second_upper_estimate": 0.00401831242747314
+      "required_token_positions_per_second": 9663.659437556054,
+      "required_pages_per_second_upper_estimate": 0.0040459724063482035
     },
     {
       "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": 19195.18914754441,
-      "required_pages_per_second_upper_estimate": 0.003997851688254453
+      "required_token_positions_per_second": 19327.31887511211,
+      "required_pages_per_second_upper_estimate": 0.004025370826011566
     },
     {
       "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": 28792.783721316613,
-      "required_pages_per_second_upper_estimate": 0.0017260834200577887
+      "required_token_positions_per_second": 28990.97831266816,
+      "required_pages_per_second_upper_estimate": 0.0017379648831836957
     },
     {
       "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": 28792.783721316613,
-      "required_pages_per_second_upper_estimate": 0.002238512288261146
+      "required_token_positions_per_second": 28990.97831266816,
+      "required_pages_per_second_upper_estimate": 0.0022539210459728526
     },
     {
       "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": 28792.783721316613,
-      "required_pages_per_second_upper_estimate": 0.36932291429453973
+      "required_token_positions_per_second": 28990.97831266816,
+      "required_pages_per_second_upper_estimate": 0.37186514170762514
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,74 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.1442483061252915,
-    "producer_p95_seconds": 0.33357249200344086,
-    "gpu_update_seconds": 0.9602406029095034,
-    "producer_batches_per_second": 6.932490417817578,
-    "gpu_batches_per_second": 1.041405661216602,
-    "producer_headroom": 6.656858778469507,
+    "producer_batch_seconds": 0.05302272368743388,
+    "producer_p95_seconds": 0.1645179819997793,
+    "gpu_update_seconds": 0.9536759919522508,
+    "producer_batches_per_second": 18.859838394854,
+    "gpu_batches_per_second": 1.048574157720926,
+    "producer_headroom": 17.98617508927153,
     "sustainable": true,
-    "buffer_seconds": 60.49515798329872,
+    "buffer_seconds": 60.0815874929918,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 87,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 247072.63852710513,
-    "upload_bytes_per_second": 5145160.357073876,
+    "generation_bytes_per_second": 433303.45547621214,
+    "upload_bytes_per_second": 3085286.3142732033,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 53.01603667125573,
+    "estimated_full_shards_per_hour": 92.97683475699209,
     "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": 1.046063713998592,
-      "bytes": 5944086,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.323775843000476,
-      "bytes": 7235009,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.370845877998363,
-      "bytes": 11121979,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.6951102359998913,
-      "bytes": 7495029,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1257126100026653,
-      "bytes": 946737,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3359714920006809,
-      "bytes": 10736001,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2456738490000134,
-      "bytes": 7708205,
-      "pages": 43,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1626391710015014,
-      "bytes": 1209462,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3864067120011896,
-      "bytes": 11713877,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.506981171998632,
-      "bytes": 11206977,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8770163509980193,
-      "bytes": 891965,
-      "pages": 27,
+      "seconds": 1.8764951709999877,
+      "bytes": 52123,
+      "pages": 1,
       "shards": 1
     },
     {
-      "seconds": 1.4873648249995313,
-      "bytes": 7797403,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1096691220009234,
-      "bytes": 969203,
-      "pages": 31,
-      "shards": 1
-    },
-    {
-      "seconds": 1.6109700509987306,
-      "bytes": 10483727,
-      "pages": 33,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5793520259976503,
-      "bytes": 13424823,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9566230939963134,
-      "bytes": 1183559,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4667487440019613,
-      "bytes": 7634603,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2097738270022091,
-      "bytes": 9815120,
+      "seconds": 1.2044020090006597,
+      "bytes": 7888698,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.1665725029961322,
-      "bytes": 1196873,
-      "pages": 37,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1840190020011505,
-      "bytes": 7418937,
+      "seconds": 1.2931400809993647,
+      "bytes": 6805061,
       "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9838692140037892,
-      "bytes": 6891211,
-      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.3423998879952705,
-      "bytes": 11990725,
-      "pages": 41,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0155430449958658,
-      "bytes": 5875734,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 2.4478168570058187,
-      "bytes": 1176227,
-      "pages": 37,
-      "shards": 1
-    },
-    {
-      "seconds": 1.676982698001666,
-      "bytes": 6919911,
-      "pages": 22,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0402027899981476,
-      "bytes": 5156086,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5196284359990386,
-      "bytes": 12048583,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2240972300060093,
-      "bytes": 7561561,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9214401089993771,
-      "bytes": 1231310,
+      "seconds": 1.581821284000398,
+      "bytes": 6273764,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.6782647159998305,
-      "bytes": 12065171,
-      "pages": 33,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0347107170018717,
-      "bytes": 6246238,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.040108234003128,
-      "bytes": 901723,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.5148740549993818,
-      "bytes": 14069210,
-      "pages": 39,
-      "shards": 4
-    },
-    {
-      "seconds": 1.4405004430009285,
-      "bytes": 5051876,
+      "seconds": 0.9842661719994794,
+      "bytes": 2388376,
       "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.163049670001783,
-      "bytes": 6248444,
+      "seconds": 4.260054792999654,
+      "bytes": 10427634,
       "pages": 39,
-      "shards": 2
-    },
-    {
-      "seconds": 3.389049104996957,
-      "bytes": 2610225,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3562926899976446,
-      "bytes": 6176571,
-      "pages": 34,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3079960410032072,
-      "bytes": 12000675,
-      "pages": 34,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1082588089993806,
-      "bytes": 7050197,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0682790589999058,
-      "bytes": 6087723,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1598256950019277,
-      "bytes": 7463538,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2493024350042106,
-      "bytes": 5521564,
-      "pages": 41,
-      "shards": 2
-    },
-    {
-      "seconds": 1.7268630670005223,
-      "bytes": 7581932,
-      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.9802855540037854,
-      "bytes": 1121294,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 2.06611223800428,
-      "bytes": 11832847,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2940340749992174,
-      "bytes": 12187860,
+      "seconds": 1.4442174710002291,
+      "bytes": 7300139,
       "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0110984060011106,
-      "bytes": 1179043,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.417530385995633,
-      "bytes": 5826314,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3201328729992383,
-      "bytes": 7683555,
-      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.031134179997025,
-      "bytes": 5438180,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.582061741995858,
-      "bytes": 13195042,
-      "pages": 34,
-      "shards": 4
-    },
-    {
-      "seconds": 0.8400354289988172,
-      "bytes": 1240516,
-      "pages": 36,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1406454610041692,
-      "bytes": 5969910,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.6059057809980004,
-      "bytes": 11827384,
-      "pages": 24,
-      "shards": 4
-    },
-    {
-      "seconds": 1.091920383994875,
-      "bytes": 1379660,
-      "pages": 39,
-      "shards": 1
-    },
-    {
-      "seconds": 1.415137545001926,
-      "bytes": 7393395,
+      "seconds": 1.0513844460001565,
+      "bytes": 1119612,
       "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3102802950015757,
-      "bytes": 11967731,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9184283650029101,
-      "bytes": 675909,
-      "pages": 20,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4080826390054426,
-      "bytes": 11404109,
-      "pages": 41,
-      "shards": 5
-    },
-    {
-      "seconds": 1.4213992930017412,
-      "bytes": 1225275,
-      "pages": 36,
       "shards": 1
-    },
-    {
-      "seconds": 1.2654285969983903,
-      "bytes": 6147750,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2639681400032714,
-      "bytes": 11134689,
-      "pages": 41,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0753672879945952,
-      "bytes": 6680589,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2491716029981035,
-      "bytes": 7586487,
-      "pages": 41,
-      "shards": 3
     }
   ],
   "cache_key": {
@@ -606,7 +270,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    2.975958035000019,
-    0.01771132999999736
+    3.415684295999995,
+    0.013631696000004467
   ]
 }

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520
-size 721778943
+oid sha256:c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56
+size 721765823