Qwen90M step 515000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 510000,
+  "step": 515000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61",
+    "model.safetensors": "a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c",
+    "trainer.pt": "ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22",
     "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
     "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
-    "shard_scheduler.json": "8e1dcb9a6b41bfa5fb3adfa01b8d6110beb427eb27007634bc7f22a838f0dfc5",
-    "coverage.json": "d8336715320aedee9a1a94d413b7034cb41f2357aa8f3ab987da371fbc49a337",
+    "shard_scheduler.json": "ba84c281f6b706297b3d60285a57ade76d674a22362d7ffd41755519d1b1e1eb",
+    "coverage.json": "ab9b69f5c3f1e54e5942d2dbadb0d1da6152c604c5616a3c283412918e4e3e24",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 510000,
+  "step": 515000,
   "stats": {
-    "climbmix/blocks": 18360000,
-    "climbmix/targets": 30689271681,
-    "climbmix/input_tokens": 30691216792,
-    "ultra_qa/blocks": 6885000,
-    "ultra_qa/targets": 11231922686,
-    "ultra_qa/input_tokens": 11231944353,
-    "cortex/blocks": 6885000,
-    "cortex/targets": 261847115,
-    "cortex/input_tokens": 10944570168,
-    "ultra_style/blocks": 6885000,
-    "ultra_style/targets": 10889862082,
-    "ultra_style/input_tokens": 10890432997,
-    "openmath/blocks": 4590000,
-    "openmath/targets": 7092338265,
-    "openmath/input_tokens": 7094701522,
-    "rewrite6/blocks": 2295000,
-    "rewrite6/targets": 3762269811,
-    "rewrite6/input_tokens": 3766394881
+    "climbmix/blocks": 18540000,
+    "climbmix/targets": 30990604912,
+    "climbmix/input_tokens": 30992568790,
+    "ultra_qa/blocks": 6952500,
+    "ultra_qa/targets": 11342093677,
+    "ultra_qa/input_tokens": 11342115570,
+    "cortex/blocks": 6952500,
+    "cortex/targets": 264409093,
+    "cortex/input_tokens": 11051859128,
+    "ultra_style/blocks": 6952500,
+    "ultra_style/targets": 10996098074,
+    "ultra_style/input_tokens": 10996674542,
+    "openmath/blocks": 4635000,
+    "openmath/targets": 7161856281,
+    "openmath/input_tokens": 7164242696,
+    "rewrite6/blocks": 2317500,
+    "rewrite6/targets": 3799257439,
+    "rewrite6/input_tokens": 3803422876
   },
   "sources": {
     "climbmix": {
-      "page": 2515,
-      "row": 756,
+      "page": 2539,
+      "row": 2373,
       "start": {
-        "file": 121,
+        "file": 122,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 56,
+          "row_group": 70,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 10302196
+      "documents": 10402117
     },
     "rewrite6": {
-      "page": 30,
-      "row": 2578,
+      "page": 44,
+      "row": 2713,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 122880
+          "row": 180224
         }
       },
       "epoch": 28,
       "offset": 0,
-      "documents": 5876770
+      "documents": 5934249
     },
     "openmath": {
-      "page": 1571,
-      "row": 2225,
+      "page": 1586,
+      "row": 3895,
       "start": {
         "file": 14,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 321,
-          "row": 716
+          "row_group": 383,
+          "row": 156
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 6437041
+      "documents": 6500151
     },
     "ultra_qa": {
-      "page": 679,
-      "row": 1959,
+      "page": 686,
+      "row": 590,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 2,
-          "row": 54023
+          "row_group": 3,
+          "row": 18153
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2783143
+      "offset": 4096,
+      "documents": 2810446
     },
     "ultra_style": {
-      "page": 1036,
-      "row": 3816,
+      "page": 1047,
+      "row": 806,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 1,
-          "row": 48587
+          "row_group": 2,
+          "row": 8236
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 4247272
+      "documents": 4289318
     },
     "cortex": {
-      "page": 133533,
-      "row": 97,
+      "page": 134842,
+      "row": 79,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 17092321
+      "documents": 17259855
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 510000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "8e1dcb9a6b41bfa5fb3adfa01b8d6110beb427eb27007634bc7f22a838f0dfc5", "coverage.json": "d8336715320aedee9a1a94d413b7034cb41f2357aa8f3ab987da371fbc49a337", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000510000"}
+{"step": 515000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ba84c281f6b706297b3d60285a57ade76d674a22362d7ffd41755519d1b1e1eb", "coverage.json": "ab9b69f5c3f1e54e5942d2dbadb0d1da6152c604c5616a3c283412918e4e3e24", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000515000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61
+oid sha256:a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892
 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": 75832.22366015913,
-      "required_pages_per_second_upper_estimate": 0.005246249187810568
+      "required_token_positions_per_second": 75955.3176128496,
+      "required_pages_per_second_upper_estimate": 0.005254765113075015
     },
     {
       "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": 9479.027957519891,
-      "required_pages_per_second_upper_estimate": 0.003968671061200793
+      "required_token_positions_per_second": 9494.4147016062,
+      "required_pages_per_second_upper_estimate": 0.003975113169637976
     },
     {
       "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": 18958.055915039782,
-      "required_pages_per_second_upper_estimate": 0.003948463089547618
+      "required_token_positions_per_second": 18988.8294032124,
+      "required_pages_per_second_upper_estimate": 0.003954872395582516
     },
     {
       "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": 28437.083872559673,
-      "required_pages_per_second_upper_estimate": 0.0017047597572470315
+      "required_token_positions_per_second": 28483.244104818597,
+      "required_pages_per_second_upper_estimate": 0.0017075269926883603
     },
     {
       "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": 28437.083872559673,
-      "required_pages_per_second_upper_estimate": 0.002210858189578581
+      "required_token_positions_per_second": 28483.244104818597,
+      "required_pages_per_second_upper_estimate": 0.002214446944599308
     },
     {
       "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": 22,
-      "required_token_positions_per_second": 28437.083872559673,
-      "required_pages_per_second_upper_estimate": 0.36476037855542737
+      "required_token_positions_per_second": 28483.244104818597,
+      "required_pages_per_second_upper_estimate": 0.365352472451849
     }
   ],
   "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": 704566,
-    "producer_batch_seconds": 0.08594841671902032,
-    "producer_p95_seconds": 0.3017171040009998,
-    "gpu_update_seconds": 0.9722515896462541,
-    "producer_batches_per_second": 11.63488564622623,
-    "gpu_batches_per_second": 1.0285403599739467,
-    "producer_headroom": 11.312036064895837,
+    "producer_batch_seconds": 0.3101006119218823,
+    "producer_p95_seconds": 0.3071302360003756,
+    "gpu_update_seconds": 0.9706759489282579,
+    "producer_batches_per_second": 3.2247598410154406,
+    "gpu_batches_per_second": 1.0302099285597004,
+    "producer_headroom": 3.130196818743401,
     "sustainable": true,
-    "buffer_seconds": 60.279598558067754,
+    "buffer_seconds": 60.181908833551994,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 339437.1945102421,
-    "upload_bytes_per_second": 6151997.828176973,
+    "generation_bytes_per_second": 525135.3264741183,
+    "upload_bytes_per_second": 6072452.474192158,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 72.83532024841735,
+    "estimated_full_shards_per_hour": 112.68181653659498,
     "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": 0.7940784490019723,
-      "bytes": 968621,
-      "pages": 29,
+      "seconds": 0.9707876079992275,
+      "bytes": 899756,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.324730037998961,
-      "bytes": 13434552,
-      "pages": 42,
-      "shards": 4
-    },
-    {
-      "seconds": 1.516792244001408,
-      "bytes": 10349687,
-      "pages": 29,
+      "seconds": 1.1750758769994718,
+      "bytes": 10630556,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.0428093720001925,
-      "bytes": 2636885,
+      "seconds": 1.1581052289984655,
+      "bytes": 2607511,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.006605718001083,
-      "bytes": 6127707,
-      "pages": 28,
+      "seconds": 0.9523471370011976,
+      "bytes": 5723812,
+      "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 1.116981245999341,
-      "bytes": 6115742,
+      "seconds": 1.186567714001285,
+      "bytes": 11920154,
+      "pages": 44,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9970581140005379,
+      "bytes": 6389566,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1336718419988756,
+      "bytes": 6075801,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.009036264000315,
-      "bytes": 6506928,
-      "pages": 34,
+      "seconds": 1.0171541380004783,
+      "bytes": 7685246,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.1265610020018357,
-      "bytes": 11819263,
+      "seconds": 0.8166221829997085,
+      "bytes": 1223594,
       "pages": 35,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.0389532609988237,
-      "bytes": 7773400,
-      "pages": 39,
+      "seconds": 1.285200268997869,
+      "bytes": 10185249,
+      "pages": 22,
       "shards": 3
     },
     {
-      "seconds": 0.9040890260002925,
-      "bytes": 1006858,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.1744725909993576,
+      "bytes": 8758656,
+      "pages": 47,
+      "shards": 3
     },
     {
-      "seconds": 1.4120586079989152,
-      "bytes": 10016675,
+      "seconds": 1.0896003289999499,
+      "bytes": 6195866,
       "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1199947279965272,
+      "bytes": 7293697,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0292591729994456,
-      "bytes": 2310644,
-      "pages": 28,
+      "seconds": 1.0610679239980527,
+      "bytes": 5725336,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.012728962003166,
-      "bytes": 6259160,
+      "seconds": 0.9264001560004544,
+      "bytes": 6158826,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9536746940029843,
-      "bytes": 6477995,
-      "pages": 23,
-      "shards": 2
+      "seconds": 1.2382891359993664,
+      "bytes": 13055404,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.2943796040017332,
-      "bytes": 12517532,
-      "pages": 41,
-      "shards": 4
+      "seconds": 0.9811806699981389,
+      "bytes": 1058761,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.182365384996956,
-      "bytes": 6177943,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.215036166999198,
+      "bytes": 9780503,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 0.8839036259996647,
-      "bytes": 2483088,
-      "pages": 29,
+      "seconds": 0.9811340400010522,
+      "bytes": 2572134,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.165388396999333,
-      "bytes": 6302666,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.036350621998281,
+      "bytes": 6380141,
+      "pages": 38,
+      "shards": 2
     },
     {
-      "seconds": 1.2500933689989324,
-      "bytes": 4989357,
-      "pages": 30,
+      "seconds": 0.9442703610002354,
+      "bytes": 2120655,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.2095713759990758,
-      "bytes": 7770394,
-      "pages": 34,
+      "seconds": 1.0825822449987754,
+      "bytes": 10431134,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.1173389949981356,
-      "bytes": 12086613,
-      "pages": 41,
+      "seconds": 1.0676360670004215,
+      "bytes": 12466956,
+      "pages": 47,
       "shards": 3
     },
     {
-      "seconds": 0.9005163840010937,
-      "bytes": 1099069,
-      "pages": 30,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1803624109998054,
-      "bytes": 11418078,
-      "pages": 40,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0558311360000516,
-      "bytes": 6077678,
-      "pages": 30,
+      "seconds": 0.9538209790007386,
+      "bytes": 2685432,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.0529501199998776,
-      "bytes": 2460568,
-      "pages": 35,
+      "seconds": 0.9428314830001909,
+      "bytes": 5761904,
+      "pages": 20,
       "shards": 2
     },
     {
-      "seconds": 1.1411234580009477,
-      "bytes": 10341495,
-      "pages": 29,
+      "seconds": 1.1127153689994884,
+      "bytes": 6912505,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.1926653939990501,
-      "bytes": 7075984,
-      "pages": 38,
+      "seconds": 1.2596074149987544,
+      "bytes": 6204319,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1058772389988007,
-      "bytes": 7919154,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.395315304001997,
-      "bytes": 5693810,
-      "pages": 28,
+      "seconds": 1.0030813440025668,
+      "bytes": 5714103,
+      "pages": 25,
       "shards": 2
     },
     {
-      "seconds": 1.0359600679985306,
-      "bytes": 6649243,
-      "pages": 34,
+      "seconds": 1.237253362996853,
+      "bytes": 8511115,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9754138270000112,
-      "bytes": 5934822,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2614193410008738,
+      "bytes": 9853756,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 0.7901811400006409,
-      "bytes": 1002646,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.0838980580010684,
+      "bytes": 8174651,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.1965479969985608,
-      "bytes": 13159361,
-      "pages": 38,
-      "shards": 4
+      "seconds": 1.1033362929993018,
+      "bytes": 798343,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.0418553449999308,
-      "bytes": 11343588,
-      "pages": 42,
+      "seconds": 1.2149433689992293,
+      "bytes": 11402427,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 0.967730065996875,
-      "bytes": 2342541,
-      "pages": 28,
+      "seconds": 0.7934381430022768,
+      "bytes": 2164331,
+      "pages": 22,
       "shards": 2
     },
     {
-      "seconds": 1.1295750149984087,
-      "bytes": 5944567,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.25598544499735,
+      "bytes": 11955544,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.1342636209992634,
-      "bytes": 6342296,
-      "pages": 37,
+      "seconds": 1.0177433429998928,
+      "bytes": 6488753,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.3789669100005995,
-      "bytes": 6252708,
-      "pages": 25,
-      "shards": 3
+      "seconds": 0.9433292390021961,
+      "bytes": 2349435,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.107486043998506,
-      "bytes": 11895475,
+      "seconds": 1.25489443799961,
+      "bytes": 9999755,
       "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.5734916069995961,
-      "bytes": 7449351,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9707876079992275,
-      "bytes": 899756,
+      "seconds": 0.9152432709997811,
+      "bytes": 861201,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.1750758769994718,
-      "bytes": 10630556,
-      "pages": 35,
+      "seconds": 1.3237921879990608,
+      "bytes": 13505764,
+      "pages": 42,
+      "shards": 5
+    },
+    {
+      "seconds": 1.3861956760010798,
+      "bytes": 10089467,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.1581052289984655,
-      "bytes": 2607511,
-      "pages": 37,
+      "seconds": 1.1900551170001563,
+      "bytes": 2519802,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 0.9523471370011976,
-      "bytes": 5723812,
-      "pages": 24,
+      "seconds": 1.0123512749996735,
+      "bytes": 5925157,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.186567714001285,
-      "bytes": 11920154,
-      "pages": 44,
-      "shards": 3
+      "seconds": 0.84063605999836,
+      "bytes": 1225280,
+      "pages": 36,
+      "shards": 1
     },
     {
-      "seconds": 0.9970581140005379,
-      "bytes": 6389566,
-      "pages": 36,
+      "seconds": 1.2510575380001683,
+      "bytes": 12042377,
+      "pages": 34,
+      "shards": 4
+    },
+    {
+      "seconds": 1.23156422700049,
+      "bytes": 11911377,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.1336718419988756,
-      "bytes": 6075801,
+      "seconds": 0.8838252950008609,
+      "bytes": 2597092,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0171541380004783,
-      "bytes": 7685246,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8166221829997085,
-      "bytes": 1223594,
-      "pages": 35,
+      "seconds": 1.0776322439996875,
+      "bytes": 6282866,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.285200268997869,
-      "bytes": 10185249,
-      "pages": 22,
-      "shards": 3
+      "seconds": 1.0567811459986842,
+      "bytes": 4557138,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.1744725909993576,
-      "bytes": 8758656,
-      "pages": 47,
+      "seconds": 1.3797404860015376,
+      "bytes": 7238899,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.0896003289999499,
-      "bytes": 6195866,
+      "seconds": 1.0933739200008858,
+      "bytes": 6120770,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1199947279965272,
-      "bytes": 7293697,
-      "pages": 31,
-      "shards": 3
+      "seconds": 0.9898015409999061,
+      "bytes": 867435,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.0610679239980527,
-      "bytes": 5725336,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.4056065479999234,
+      "bytes": 17665783,
+      "pages": 34,
+      "shards": 5
     },
     {
-      "seconds": 0.9264001560004544,
-      "bytes": 6158826,
+      "seconds": 1.0921460530007607,
+      "bytes": 6276750,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2382891359993664,
-      "bytes": 13055404,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0008894900020096,
+      "bytes": 2319506,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 0.9811806699981389,
-      "bytes": 1058761,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.162798567002028,
+      "bytes": 6350229,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.215036166999198,
-      "bytes": 9780503,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.0111840479985403,
+      "bytes": 5239532,
+      "pages": 38,
+      "shards": 2
     },
     {
-      "seconds": 0.9811340400010522,
-      "bytes": 2572134,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2082140299971798,
+      "bytes": 13368136,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.036350621998281,
-      "bytes": 6380141,
-      "pages": 38,
+      "seconds": 1.2696354079998855,
+      "bytes": 6070118,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.9442703610002354,
-      "bytes": 2120655,
-      "pages": 20,
+      "seconds": 0.9898074049997376,
+      "bytes": 2654933,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.0825822449987754,
-      "bytes": 10431134,
-      "pages": 32,
+      "seconds": 1.3211888400001044,
+      "bytes": 9995304,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0676360670004215,
-      "bytes": 12466956,
-      "pages": 47,
+      "seconds": 0.8846473650010012,
+      "bytes": 1315669,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.6196549160013092,
+      "bytes": 7467278,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 0.9538209790007386,
-      "bytes": 2685432,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.2983564580026723,
+      "bytes": 12027088,
+      "pages": 34,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c
-size 721759487
+oid sha256:ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22
+size 721771839