Qwen90M step 530000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 525000,
+  "step": 530000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c",
+    "model.safetensors": "3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da",
+    "trainer.pt": "4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35",
     "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
-    "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
-    "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553",
-    "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b",
+    "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
+    "shard_scheduler.json": "3e8b88313d070ffdb6572b9af938d708afe66786c34590df9f577e00e5a7ad71",
+    "coverage.json": "1f166ed6ce6875dfc663aa1359c72063b2c84099952c88e471aa37b08f50927f",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 525000,
+  "step": 530000,
   "stats": {
-    "climbmix/blocks": 18900000,
-    "climbmix/targets": 31592797042,
-    "climbmix/input_tokens": 31594799016,
-    "ultra_qa/blocks": 7087500,
-    "ultra_qa/targets": 11562286610,
-    "ultra_qa/input_tokens": 11562308928,
-    "cortex/blocks": 7087500,
-    "cortex/targets": 269544792,
-    "cortex/input_tokens": 11266328102,
-    "ultra_style/blocks": 7087500,
-    "ultra_style/targets": 11209247634,
-    "ultra_style/input_tokens": 11209834992,
-    "openmath/blocks": 4725000,
-    "openmath/targets": 7301172470,
-    "openmath/input_tokens": 7303605185,
-    "rewrite6/blocks": 2362500,
-    "rewrite6/targets": 3872985034,
-    "rewrite6/input_tokens": 3877231666
+    "climbmix/blocks": 19080000,
+    "climbmix/targets": 31893841018,
+    "climbmix/input_tokens": 31895861954,
+    "ultra_qa/blocks": 7155000,
+    "ultra_qa/targets": 11672418639,
+    "ultra_qa/input_tokens": 11672441161,
+    "cortex/blocks": 7155000,
+    "cortex/targets": 272094364,
+    "cortex/input_tokens": 11373535505,
+    "ultra_style/blocks": 7155000,
+    "ultra_style/targets": 11316260253,
+    "ultra_style/input_tokens": 11316853356,
+    "openmath/blocks": 4770000,
+    "openmath/targets": 7370634197,
+    "openmath/input_tokens": 7373089745,
+    "rewrite6/blocks": 2385000,
+    "rewrite6/targets": 3909838802,
+    "rewrite6/input_tokens": 3914125422
   },
   "sources": {
     "climbmix": {
-      "page": 2588,
-      "row": 3394,
+      "page": 2613,
+      "row": 1722,
       "start": {
-        "file": 125,
+        "file": 126,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 20,
+          "row_group": 37,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 10603842
+      "offset": 0,
+      "documents": 10704570
     },
     "rewrite6": {
-      "page": 22,
-      "row": 2978,
+      "page": 36,
+      "row": 2819,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 90112
+          "row": 147456
         }
       },
       "epoch": 29,
       "offset": 0,
-      "documents": 6049806
+      "documents": 6106991
     },
     "openmath": {
-      "page": 1617,
-      "row": 2871,
+      "page": 1633,
+      "row": 76,
       "start": {
         "file": 15,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 73,
-          "row": 482
+          "row_group": 139,
+          "row": 18
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 6626103
+      "documents": 6688844
     },
     "ultra_qa": {
-      "page": 699,
-      "row": 1991,
+      "page": 706,
+      "row": 539,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 4,
-          "row": 6884
+          "row": 35556
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2865095
+      "documents": 2892315
     },
     "ultra_style": {
-      "page": 1067,
-      "row": 1860,
+      "page": 1077,
+      "row": 2689,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 3,
-          "row": 4718
+          "row": 45678
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4372292
+      "documents": 4414081
     },
     "cortex": {
-      "page": 137458,
-      "row": 115,
+      "page": 138765,
+      "row": 110,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 17594739
+      "documents": 17762030
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 525000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553", "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000525000"}
+{"step": 530000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "3e8b88313d070ffdb6572b9af938d708afe66786c34590df9f577e00e5a7ad71", "coverage.json": "1f166ed6ce6875dfc663aa1359c72063b2c84099952c88e471aa37b08f50927f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000530000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c
+oid sha256:3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189
 size 360622744

performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 90,
     "mode": "compiled",
-    "microbatch_seconds": 0.9387972919998901,
+    "microbatch_seconds": 0.9447870520000379,
     "peak_gb": 89.37766551971436,
     "fits_budget": true,
-    "estimated_update_seconds": 0.9387972919998901,
-    "input_tokens_per_second": 196336.31410178967
+    "estimated_update_seconds": 0.9447870520000379,
+    "input_tokens_per_second": 195091.58133550777
   },
   "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": 77213.6919826638,
-      "required_pages_per_second_upper_estimate": 0.005341822371809575
+      "required_token_positions_per_second": 77015.98464267456,
+      "required_pages_per_second_upper_estimate": 0.005328144519284888
     },
     {
       "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": 9651.711497832976,
-      "required_pages_per_second_upper_estimate": 0.004040970053487503
+      "required_token_positions_per_second": 9626.99808033432,
+      "required_pages_per_second_upper_estimate": 0.0040306230616556585
     },
     {
       "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": 19303.42299566595,
-      "required_pages_per_second_upper_estimate": 0.00402039394449965
+      "required_token_positions_per_second": 19253.99616066864,
+      "required_pages_per_second_upper_estimate": 0.004010099638243004
     },
     {
       "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": 28955.13449349893,
-      "required_pages_per_second_upper_estimate": 0.001735816100954843
+      "required_token_positions_per_second": 28880.99424100296,
+      "required_pages_per_second_upper_estimate": 0.0017313715060233205
     },
     {
       "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": 28955.13449349893,
-      "required_pages_per_second_upper_estimate": 0.002251134346692743
+      "required_token_positions_per_second": 28880.99424100296,
+      "required_pages_per_second_upper_estimate": 0.0022453702681696883
     },
     {
       "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": 28955.13449349893,
-      "required_pages_per_second_upper_estimate": 0.37140537568141674
+      "required_token_positions_per_second": 28880.99424100296,
+      "required_pages_per_second_upper_estimate": 0.3704543841279994
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,182 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.11904607814852852,
-    "producer_p95_seconds": 0.3377903690052335,
-    "gpu_update_seconds": 0.9548565559661825,
-    "producer_batches_per_second": 8.400108727247144,
-    "gpu_batches_per_second": 1.0472777232891683,
-    "producer_headroom": 8.02089888904068,
+    "producer_batch_seconds": 0.320096299234379,
+    "producer_p95_seconds": 0.15855196899974544,
+    "gpu_update_seconds": 0.95730776334381,
+    "producer_batches_per_second": 3.1240598607101857,
+    "gpu_batches_per_second": 1.0445961458696094,
+    "producer_headroom": 2.9906867578086427,
     "sustainable": true,
-    "buffer_seconds": 60.1559630258695,
+    "buffer_seconds": 60.310389090660024,
     "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": 384588.48339903663,
-    "upload_bytes_per_second": 3837530.023601385,
+    "generation_bytes_per_second": 386229.8712076355,
+    "upload_bytes_per_second": 3476176.657096584,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 82.52373577574086,
+    "estimated_full_shards_per_hour": 82.87593938991354,
     "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": 1.521548144002736,
-      "bytes": 10001036,
-      "pages": 42,
-      "shards": 3
-    },
-    {
-      "seconds": 1.037408388001495,
-      "bytes": 2352962,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4666764929934288,
-      "bytes": 11848768,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1809732929978054,
-      "bytes": 12035253,
-      "pages": 37,
-      "shards": 4
-    },
-    {
-      "seconds": 0.976799342999584,
-      "bytes": 895629,
-      "pages": 27,
+      "seconds": 1.0629590909998115,
+      "bytes": 25106,
+      "pages": 1,
       "shards": 1
     },
     {
-      "seconds": 1.1010828580037924,
-      "bytes": 6038977,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0995818070005043,
-      "bytes": 2583346,
+      "seconds": 1.218460234000304,
+      "bytes": 6273746,
       "pages": 34,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 0.9759389660030138,
-      "bytes": 6031085,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.1453777450001326,
+      "bytes": 12048299,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.012582071001816,
-      "bytes": 5087118,
-      "pages": 25,
-      "shards": 2
+      "seconds": 1.011093388000063,
+      "bytes": 1199201,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.1673960980042466,
-      "bytes": 13547167,
-      "pages": 41,
+      "seconds": 1.0240878049999083,
+      "bytes": 7205273,
+      "pages": 28,
       "shards": 4
     },
     {
-      "seconds": 1.3716537860018434,
-      "bytes": 6334955,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.127826156996889,
-      "bytes": 2574706,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1363400170012028,
-      "bytes": 10297243,
-      "pages": 34,
+      "seconds": 1.1896957620001558,
+      "bytes": 10758587,
+      "pages": 42,
       "shards": 3
     },
     {
-      "seconds": 1.2236347540019779,
-      "bytes": 6431007,
-      "pages": 37,
+      "seconds": 0.9081728179999118,
+      "bytes": 2358450,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.9796522880060365,
-      "bytes": 2218113,
-      "pages": 23,
+      "seconds": 1.152849502999743,
+      "bytes": 6119990,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.3331365960038966,
-      "bytes": 12026403,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 1.336267955994117,
-      "bytes": 7188821,
-      "pages": 41,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3482005179976113,
-      "bytes": 6310167,
+      "seconds": 1.0355856849996599,
+      "bytes": 6832540,
       "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.184783762000734,
-      "bytes": 5973673,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9036034050004673,
-      "bytes": 2269487,
-      "pages": 25,
-      "shards": 2
+      "seconds": 20.55054765299974,
+      "bytes": 11612692,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.6477199560031295,
-      "bytes": 16059670,
+      "seconds": 1.2162191410006926,
+      "bytes": 6401722,
       "pages": 45,
-      "shards": 5
-    },
-    {
-      "seconds": 1.0709953250043327,
-      "bytes": 6125208,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8623744039941812,
-      "bytes": 2649650,
-      "pages": 39,
       "shards": 2
     },
     {
-      "seconds": 1.4692237859999295,
-      "bytes": 10546134,
-      "pages": 31,
+      "seconds": 1.3737986060004914,
+      "bytes": 7819084,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1281011890023365,
-      "bytes": 2729862,
-      "pages": 38,
+      "seconds": 1.079906239000593,
+      "bytes": 5671187,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 2.671323365997523,
-      "bytes": 5596441,
-      "pages": 20,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0342092059945571,
-      "bytes": 930671,
-      "pages": 29,
-      "shards": 1
-    },
-    {
-      "seconds": 1.692445526998199,
-      "bytes": 13078461,
-      "pages": 32,
-      "shards": 4
-    },
-    {
-      "seconds": 1.3595879119966412,
-      "bytes": 10348219,
-      "pages": 44,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0515126419995795,
-      "bytes": 1003171,
-      "pages": 30,
-      "shards": 1
-    },
-    {
-      "seconds": 1.000422951001383,
-      "bytes": 7649877,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2835066039988305,
-      "bytes": 10437972,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8616337460043724,
-      "bytes": 2430044,
+      "seconds": 1.1921040119996178,
+      "bytes": 5850261,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3565980959974695,
-      "bytes": 12199549,
-      "pages": 43,
+      "seconds": 0.9688384539995241,
+      "bytes": 8315017,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.012810416999855,
-      "bytes": 6201884,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8832714039963321,
-      "bytes": 2126873,
-      "pages": 22,
+      "seconds": 1.0787890209994657,
+      "bytes": 6174644,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.4999150360017666,
-      "bytes": 10166456,
-      "pages": 43,
+      "seconds": 1.431994874999873,
+      "bytes": 7632897,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9999620900052832,
-      "bytes": 2535870,
-      "pages": 34,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0642618760030018,
-      "bytes": 5883611,
-      "pages": 24,
-      "shards": 2
-    },
-    {
-      "seconds": 2.4567707870010054,
-      "bytes": 16775004,
-      "pages": 44,
-      "shards": 4
-    },
-    {
-      "seconds": 1.3269825160023174,
-      "bytes": 2735395,
-      "pages": 38,
+      "seconds": 1.237401497999599,
+      "bytes": 5052734,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 2.3559383290048572,
-      "bytes": 6286319,
-      "pages": 38,
+      "seconds": 1.0236900689997128,
+      "bytes": 6421100,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9432465139980195,
-      "bytes": 858963,
+      "seconds": 0.9477791999997862,
+      "bytes": 2414897,
       "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.6026845509986742,
-      "bytes": 11607409,
-      "pages": 29,
-      "shards": 4
-    },
-    {
-      "seconds": 1.04873497600056,
-      "bytes": 6766779,
-      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.4542191169966827,
-      "bytes": 7891732,
-      "pages": 46,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0971116009968682,
-      "bytes": 6168592,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.321997156002908,
-      "bytes": 9700555,
-      "pages": 34,
+      "seconds": 1.0987587250001525,
+      "bytes": 11899150,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.1995469100002083,
-      "bytes": 2400407,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5711508849999518,
-      "bytes": 6147230,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4379745080004795,
-      "bytes": 8148657,
-      "pages": 31,
+      "seconds": 1.4687352659993849,
+      "bytes": 10700369,
+      "pages": 42,
       "shards": 3
     },
     {
-      "seconds": 1.2602131539970287,
-      "bytes": 10952650,
-      "pages": 39,
-      "shards": 4
-    },
-    {
-      "seconds": 1.072258756001247,
-      "bytes": 6000592,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8994364020036301,
-      "bytes": 2733894,
+      "seconds": 1.1858354729993152,
+      "bytes": 2742758,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 3.8195672939982614,
-      "bytes": 6121969,
-      "pages": 34,
+      "seconds": 1.046531567000784,
+      "bytes": 5904437,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2331268449997879,
-      "bytes": 7295036,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 31.867704994001542,
-      "bytes": 12094685,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.475563729996793,
-      "bytes": 6268574,
+      "seconds": 1.4689589510007863,
+      "bytes": 7230449,
       "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4652358960011043,
-      "bytes": 7414584,
-      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 2.4988657550056814,
-      "bytes": 5294252,
-      "pages": 30,
+      "seconds": 1.0110029770003166,
+      "bytes": 6117097,
+      "pages": 37,
       "shards": 2
-    },
-    {
-      "seconds": 1.1802646279975306,
-      "bytes": 7575689,
-      "pages": 41,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8871877959973062,
-      "bytes": 955502,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.3968777099944418,
-      "bytes": 12101140,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1570743730044342,
-      "bytes": 12058095,
-      "pages": 40,
-      "shards": 4
     }
   ],
   "cache_key": {
@@ -606,7 +378,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    1.0198370660000364,
-    0.013382547000219347
+    1.0657202160000452,
+    0.013942628999984663
   ]
 }

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da
-size 721783935
+oid sha256:4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35
+size 721771007