Qwen90M step 60000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 55000,
+  "step": 60000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da",
+    "model.safetensors": "a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f",
+    "trainer.pt": "09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4",
     "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
     "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f",
-    "shard_scheduler.json": "0b7760fce951fcc476bd9ce864c73c4a8111a03bf92e84c8373e26e818a1f854",
-    "coverage.json": "a90e5fa0fd4f8822ac17717d8ddbef2645fe20188ab34ad850e1754843896bc9",
+    "shard_scheduler.json": "502ba4fb1878cbdcbe2c170d26ff10458bed187e056b37f6ad7fa2afaa31cc12",
+    "coverage.json": "558c9d8fbc6f9cdc4f9f8ba67a6152d0f15e7f96f74c310caabb313dfccd148e",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 55000,
+  "step": 60000,
   "stats": {
-    "climbmix/blocks": 1980000,
-    "climbmix/targets": 3309029909,
-    "climbmix/input_tokens": 3309239389,
-    "ultra_qa/blocks": 742500,
-    "ultra_qa/targets": 1211473004,
-    "ultra_qa/input_tokens": 1211475401,
-    "cortex/blocks": 742500,
-    "cortex/targets": 28237480,
-    "cortex/input_tokens": 1180342911,
-    "ultra_style/blocks": 742500,
-    "ultra_style/targets": 1174419713,
-    "ultra_style/input_tokens": 1174481204,
-    "openmath/blocks": 495000,
-    "openmath/targets": 764845062,
-    "openmath/input_tokens": 765100091,
-    "rewrite6/blocks": 247500,
-    "rewrite6/targets": 405729377,
-    "rewrite6/input_tokens": 406174162
+    "climbmix/blocks": 2160000,
+    "climbmix/targets": 3609663868,
+    "climbmix/input_tokens": 3609892580,
+    "ultra_qa/blocks": 810000,
+    "ultra_qa/targets": 1321742315,
+    "ultra_qa/input_tokens": 1321744914,
+    "cortex/blocks": 810000,
+    "cortex/targets": 30814012,
+    "cortex/input_tokens": 1287629180,
+    "ultra_style/blocks": 810000,
+    "ultra_style/targets": 1281312518,
+    "ultra_style/input_tokens": 1281379443,
+    "openmath/blocks": 540000,
+    "openmath/targets": 834463745,
+    "openmath/input_tokens": 834742126,
+    "rewrite6/blocks": 270000,
+    "rewrite6/targets": 442578795,
+    "rewrite6/input_tokens": 443064258
   },
   "sources": {
     "climbmix": {
-      "page": 271,
-      "row": 208,
+      "page": 295,
+      "row": 3347,
       "start": {
-        "file": 13,
+        "file": 14,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 9,
+          "row_group": 22,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 53248,
-      "documents": 1110224
+      "offset": 2048,
+      "documents": 1211667
     },
     "rewrite6": {
-      "page": 4,
-      "row": 1053,
+      "page": 18,
+      "row": 1596,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 16384
+          "row": 73728
         }
       },
       "epoch": 3,
       "offset": 2048,
-      "documents": 633649
+      "documents": 691536
     },
     "openmath": {
-      "page": 169,
-      "row": 1982,
+      "page": 184,
+      "row": 3842,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 255,
-          "row": 574
+          "row_group": 317,
+          "row": 14
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 694206
+      "offset": 0,
+      "documents": 757506
     },
     "ultra_qa": {
-      "page": 73,
-      "row": 1336,
+      "page": 79,
+      "row": 4028,
       "start": {
         "file": 0,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
-          "row": 40674
+          "row_group": 5,
+          "row": 847
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 300344
+      "offset": 2048,
+      "documents": 327612
     },
     "ultra_style": {
-      "page": 111,
-      "row": 3394,
+      "page": 121,
+      "row": 3563,
       "start": {
         "file": 0,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 28852
+          "row": 69812
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 458050
+      "documents": 499179
     },
     "cortex": {
-      "page": 14397,
-      "row": 119,
+      "page": 15706,
+      "row": 87,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 1842935
+      "documents": 2010455
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 55000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "0b7760fce951fcc476bd9ce864c73c4a8111a03bf92e84c8373e26e818a1f854", "coverage.json": "a90e5fa0fd4f8822ac17717d8ddbef2645fe20188ab34ad850e1754843896bc9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000055000"}
+{"step": 60000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "502ba4fb1878cbdcbe2c170d26ff10458bed187e056b37f6ad7fa2afaa31cc12", "coverage.json": "558c9d8fbc6f9cdc4f9f8ba67a6152d0f15e7f96f74c310caabb313dfccd148e", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000060000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da
+oid sha256:a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 76127.77658254562,
-      "required_pages_per_second_upper_estimate": 0.006598562730089286
+      "required_token_positions_per_second": 75551.01885780718,
+      "required_pages_per_second_upper_estimate": 0.006548570832288048
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9515.972072818202,
-      "required_pages_per_second_upper_estimate": 0.0037960635363085216
+      "required_token_positions_per_second": 9443.877357225898,
+      "required_pages_per_second_upper_estimate": 0.0037673038763466962
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19031.944145636404,
-      "required_pages_per_second_upper_estimate": 0.004402705327316008
+      "required_token_positions_per_second": 18887.754714451796,
+      "required_pages_per_second_upper_estimate": 0.004369349640058767
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28547.916218454604,
-      "required_pages_per_second_upper_estimate": 0.0017384183183630445
+      "required_token_positions_per_second": 28331.632071677694,
+      "required_pages_per_second_upper_estimate": 0.0017252477485795504
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28547.916218454604,
-      "required_pages_per_second_upper_estimate": 0.002197697616186597
+      "required_token_positions_per_second": 28331.632071677694,
+      "required_pages_per_second_upper_estimate": 0.002181047463854873
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28547.916218454604,
-      "required_pages_per_second_upper_estimate": 0.35446957570377097
+      "required_token_positions_per_second": 28331.632071677694,
+      "required_pages_per_second_upper_estimate": 0.3517840504572767
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.09582121260153542,
-    "producer_p95_seconds": 0.27117204399837647,
-    "gpu_update_seconds": 0.9684769910501257,
-    "producer_batches_per_second": 10.436102537737831,
-    "gpu_batches_per_second": 1.0325490530401695,
-    "producer_headroom": 10.107125184038916,
+    "producer_batch_seconds": 0.09136518590602805,
+    "producer_p95_seconds": 0.29452696799853584,
+    "gpu_update_seconds": 0.9758703603820585,
+    "producer_batches_per_second": 10.945088001337089,
+    "gpu_batches_per_second": 1.0247262757406574,
+    "producer_headroom": 10.68098697227817,
     "sustainable": true,
-    "buffer_seconds": 60.0455734451078,
+    "buffer_seconds": 60.50396234368763,
     "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": 416734.52252620424,
-    "upload_bytes_per_second": 6385147.086400999,
+    "generation_bytes_per_second": 384384.7746639473,
+    "upload_bytes_per_second": 6418706.338644549,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 89.42152745093914,
+    "estimated_full_shards_per_hour": 82.48002462328733,
     "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": 1.8200730230000772,
-      "bytes": 7217811,
-      "pages": 25,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9584536009997464,
-      "bytes": 5627627,
+      "seconds": 1.1064888179989794,
+      "bytes": 5541578,
       "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.156600534000063,
-      "bytes": 13506264,
-      "pages": 39,
-      "shards": 4
+      "seconds": 0.9643621880004503,
+      "bytes": 7410728,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 0.817877588000556,
-      "bytes": 6663126,
-      "pages": 35,
+      "seconds": 0.8883736560001125,
+      "bytes": 6293157,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9133121940003548,
-      "bytes": 1030486,
-      "pages": 31,
+      "seconds": 0.9868076469992957,
+      "bytes": 927962,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.181683423999857,
-      "bytes": 12141179,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.2888746650005487,
+      "bytes": 18374728,
+      "pages": 40,
+      "shards": 5
     },
     {
-      "seconds": 1.1883426349995716,
-      "bytes": 1212298,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1437757880012214,
+      "bytes": 6056576,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.1447565099988424,
-      "bytes": 5933026,
-      "pages": 35,
+      "seconds": 1.102731920998849,
+      "bytes": 2598092,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2180850740005553,
-      "bytes": 13486469,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.1366644360005012,
+      "bytes": 5837515,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 1.0477722719988378,
-      "bytes": 4809637,
-      "pages": 36,
+      "seconds": 0.8290303609992407,
+      "bytes": 5336251,
+      "pages": 39,
       "shards": 2
     },
     {
-      "seconds": 1.1933620469990274,
-      "bytes": 7601419,
-      "pages": 37,
+      "seconds": 1.0252480340004695,
+      "bytes": 7371694,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.0362921949999873,
-      "bytes": 5998979,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.1864866850010003,
+      "bytes": 11535625,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 0.7888650649983902,
-      "bytes": 730238,
-      "pages": 23,
-      "shards": 1
+      "seconds": 0.9566465869993408,
+      "bytes": 1137762,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.1390384229998745,
-      "bytes": 12216394,
-      "pages": 44,
+      "seconds": 1.0941519280004286,
+      "bytes": 11375746,
+      "pages": 34,
       "shards": 4
     },
     {
-      "seconds": 0.9729804899998271,
-      "bytes": 6782943,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0282262380005704,
-      "bytes": 6456198,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.7203741030007222,
+      "bytes": 1052913,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.007465778999176,
-      "bytes": 7569670,
-      "pages": 36,
+      "seconds": 1.0976181589994667,
+      "bytes": 7695785,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.248841076001554,
-      "bytes": 5066284,
+      "seconds": 1.024126566000632,
+      "bytes": 5953181,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0381730779990903,
-      "bytes": 7573812,
-      "pages": 37,
+      "seconds": 1.452807390000089,
+      "bytes": 11684781,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 0.756838974000857,
-      "bytes": 1060731,
-      "pages": 33,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1045795680001902,
-      "bytes": 11876139,
-      "pages": 32,
+      "seconds": 0.8971567329990648,
+      "bytes": 7415356,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.0963092559995857,
-      "bytes": 12324311,
-      "pages": 39,
-      "shards": 5
-    },
-    {
-      "seconds": 0.7185119169989775,
-      "bytes": 873551,
-      "pages": 25,
+      "seconds": 0.9046226490008848,
+      "bytes": 1040439,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 0.8877478260001226,
-      "bytes": 6165145,
+      "seconds": 1.043625083999359,
+      "bytes": 6212867,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9784168599999248,
-      "bytes": 2018917,
-      "pages": 19,
-      "shards": 2
+      "seconds": 1.3019511900001817,
+      "bytes": 12069002,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1833946610004205,
-      "bytes": 10020680,
-      "pages": 36,
-      "shards": 3
+      "seconds": 0.8234439619991463,
+      "bytes": 1074472,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 0.9556004539990681,
-      "bytes": 6956463,
-      "pages": 35,
+      "seconds": 1.0881272910010011,
+      "bytes": 13091178,
+      "pages": 33,
+      "shards": 4
+    },
+    {
+      "seconds": 1.0808892179993563,
+      "bytes": 6450126,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.286864922998575,
-      "bytes": 7382695,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.8513057910004136,
+      "bytes": 341762,
+      "pages": 9,
+      "shards": 1
     },
     {
-      "seconds": 0.9901690320002672,
-      "bytes": 6106536,
-      "pages": 30,
+      "seconds": 1.0242165819981892,
+      "bytes": 5354140,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.1473440179997851,
-      "bytes": 6606039,
-      "pages": 35,
+      "seconds": 0.845649806999063,
+      "bytes": 8068507,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.092060394001237,
-      "bytes": 6397904,
+      "seconds": 0.9696599159997277,
+      "bytes": 6181281,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9357377939995786,
-      "bytes": 1197833,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.0534396790008032,
+      "bytes": 2069289,
+      "pages": 21,
+      "shards": 2
     },
     {
-      "seconds": 1.0255696429994714,
-      "bytes": 13323829,
-      "pages": 36,
+      "seconds": 1.112150704000669,
+      "bytes": 16576455,
+      "pages": 44,
       "shards": 4
     },
     {
-      "seconds": 1.3980831969984138,
-      "bytes": 10844649,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9555492769995908,
-      "bytes": 915953,
-      "pages": 28,
+      "seconds": 0.7744886969994695,
+      "bytes": 1014054,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.0586995369994838,
-      "bytes": 7742624,
-      "pages": 35,
+      "seconds": 0.9544328729989502,
+      "bytes": 7444287,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.8267605399996683,
-      "bytes": 1098322,
-      "pages": 34,
+      "seconds": 1.0936733489998005,
+      "bytes": 876493,
+      "pages": 25,
       "shards": 1
     },
     {
-      "seconds": 1.0487060740015295,
-      "bytes": 7770470,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.364482594999572,
+      "bytes": 12283477,
+      "pages": 38,
+      "shards": 4
     },
     {
-      "seconds": 1.5646253450013319,
-      "bytes": 15407220,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.1743630839991965,
+      "bytes": 5760431,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 0.8243736469994474,
-      "bytes": 1000268,
-      "pages": 30,
-      "shards": 1
+      "seconds": 0.9926189179968787,
+      "bytes": 7099055,
+      "pages": 43,
+      "shards": 2
     },
     {
-      "seconds": 1.0672549959999742,
-      "bytes": 7538580,
-      "pages": 38,
+      "seconds": 1.0609590189997107,
+      "bytes": 7221766,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.1064888179989794,
-      "bytes": 5541578,
-      "pages": 33,
+      "seconds": 0.9498501529997156,
+      "bytes": 6032509,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 0.9643621880004503,
-      "bytes": 7410728,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.024946067998826,
+      "bytes": 5141639,
+      "pages": 39,
+      "shards": 2
     },
     {
-      "seconds": 0.8883736560001125,
-      "bytes": 6293157,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.2369695059969672,
+      "bytes": 7463645,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 0.9868076469992957,
-      "bytes": 927962,
-      "pages": 26,
+      "seconds": 0.9742800499989244,
+      "bytes": 772407,
+      "pages": 25,
       "shards": 1
     },
     {
-      "seconds": 1.2888746650005487,
-      "bytes": 18374728,
-      "pages": 40,
+      "seconds": 1.406286874000216,
+      "bytes": 18261180,
+      "pages": 44,
       "shards": 5
     },
     {
-      "seconds": 1.1437757880012214,
-      "bytes": 6056576,
-      "pages": 36,
+      "seconds": 1.0190579750014876,
+      "bytes": 5885831,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.102731920998849,
-      "bytes": 2598092,
-      "pages": 36,
+      "seconds": 0.8101317319997179,
+      "bytes": 713378,
+      "pages": 25,
       "shards": 2
     },
     {
-      "seconds": 1.1366644360005012,
-      "bytes": 5837515,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.03739196099923,
+      "bytes": 7802740,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 0.8290303609992407,
-      "bytes": 5336251,
-      "pages": 39,
+      "seconds": 1.1620692619981128,
+      "bytes": 6079962,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.0252480340004695,
-      "bytes": 7371694,
-      "pages": 29,
-      "shards": 3
+      "seconds": 0.8016836730021168,
+      "bytes": 4650466,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.1864866850010003,
-      "bytes": 11535625,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.2261960830001044,
+      "bytes": 13273679,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 0.9566465869993408,
-      "bytes": 1137762,
-      "pages": 35,
-      "shards": 2
+      "seconds": 0.7825120570014406,
+      "bytes": 1093241,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.0941519280004286,
-      "bytes": 11375746,
-      "pages": 34,
+      "seconds": 1.4853517330011528,
+      "bytes": 11092996,
+      "pages": 24,
       "shards": 4
     },
     {
-      "seconds": 0.7203741030007222,
-      "bytes": 1052913,
+      "seconds": 0.8748619340003643,
+      "bytes": 6384371,
+      "pages": 42,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9224505880010838,
+      "bytes": 1117951,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.0976181589994667,
-      "bytes": 7695785,
-      "pages": 37,
+      "seconds": 1.0108905939996475,
+      "bytes": 7845615,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.024126566000632,
-      "bytes": 5953181,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.27624071000173,
+      "bytes": 16075425,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.452807390000089,
-      "bytes": 11684781,
-      "pages": 40,
-      "shards": 3
+      "seconds": 0.8457192399982887,
+      "bytes": 2628229,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.8971567329990648,
-      "bytes": 7415356,
-      "pages": 29,
-      "shards": 3
+      "seconds": 0.9226963619985327,
+      "bytes": 5611919,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 0.9046226490008848,
-      "bytes": 1040439,
+      "seconds": 0.7924021649996575,
+      "bytes": 1122256,
       "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.043625083999359,
-      "bytes": 6212867,
+      "seconds": 1.1671633029982331,
+      "bytes": 11363880,
+      "pages": 40,
+      "shards": 4
+    },
+    {
+      "seconds": 0.9065282010014926,
+      "bytes": 6177542,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.0971184539994283,
+      "bytes": 8482200,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1112909670009685,
+      "bytes": 6011206,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.3019511900001817,
-      "bytes": 12069002,
-      "pages": 36,
-      "shards": 4
+      "seconds": 0.9119941530007054,
+      "bytes": 5747516,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 0.8234439619991463,
-      "bytes": 1074472,
-      "pages": 32,
-      "shards": 1
+      "seconds": 1.016742787000112,
+      "bytes": 7467282,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.0881272910010011,
-      "bytes": 13091178,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.8858778819994768,
+      "bytes": 5985432,
+      "pages": 27,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f
-size 721776959
+oid sha256:09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4
+size 721780607