Qwen90M step 900000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 895000,
+  "step": 900000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd",
+    "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985",
+    "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d",
     "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
-    "shard_scheduler.json": "2d68de6c3624467095a90ec9bdeef51e2fc5ab4e71e38707f6c6cc942ef1b8c8",
-    "coverage.json": "da94cd20bec68c816560b61fce43b1062241b6873d42a2e73f505b9d916b5a6b",
+    "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9",
+    "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 895000,
+  "step": 900000,
   "stats": {
-    "climbmix/blocks": 32220000,
-    "climbmix/targets": 53861724331,
-    "climbmix/input_tokens": 53865137686,
-    "ultra_qa/blocks": 12082500,
-    "ultra_qa/targets": 19711743279,
-    "ultra_qa/input_tokens": 19711781154,
-    "cortex/blocks": 12082500,
-    "cortex/targets": 459427225,
-    "cortex/input_tokens": 19205987571,
-    "ultra_style/blocks": 12082500,
-    "ultra_style/targets": 19110329995,
-    "ultra_style/input_tokens": 19111332267,
-    "openmath/blocks": 8055000,
-    "openmath/targets": 12446359341,
-    "openmath/input_tokens": 12450506245,
-    "rewrite6/blocks": 4027500,
-    "rewrite6/targets": 6602556148,
-    "rewrite6/input_tokens": 6609794679
+    "climbmix/blocks": 32400000,
+    "climbmix/targets": 54162259750,
+    "climbmix/input_tokens": 54165692291,
+    "ultra_qa/blocks": 12150000,
+    "ultra_qa/targets": 19821847028,
+    "ultra_qa/input_tokens": 19821885096,
+    "cortex/blocks": 12150000,
+    "cortex/targets": 461992778,
+    "cortex/input_tokens": 19313376697,
+    "ultra_style/blocks": 12150000,
+    "ultra_style/targets": 19217238700,
+    "ultra_style/input_tokens": 19218246348,
+    "openmath/blocks": 8100000,
+    "openmath/targets": 12515849799,
+    "openmath/input_tokens": 12520019747,
+    "rewrite6/blocks": 4050000,
+    "rewrite6/targets": 6639430765,
+    "rewrite6/input_tokens": 6646710396
   },
   "sources": {
     "climbmix": {
-      "page": 4412,
-      "row": 2552,
+      "page": 4437,
+      "row": 1722,
       "start": {
-        "file": 213,
+        "file": 214,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 47,
+          "row_group": 65,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 18074104
+      "offset": 2048,
+      "documents": 18175674
     },
     "rewrite6": {
-      "page": 10,
-      "row": 1253,
+      "page": 24,
+      "row": 2293,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 40960
+          "row": 98304
         }
       },
       "epoch": 50,
-      "offset": 4096,
-      "documents": 10312413
+      "offset": 2048,
+      "documents": 10370797
     },
     "openmath": {
-      "page": 2757,
-      "row": 3359,
+      "page": 2773,
+      "row": 839,
       "start": {
-        "file": 25,
+        "file": 26,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 376,
-          "row": 424
+          "row_group": 5,
+          "row": 311
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 11296031
+      "offset": 2048,
+      "documents": 11359047
     },
     "ultra_qa": {
-      "page": 1192,
-      "row": 1692,
+      "page": 1199,
+      "row": 201,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 3,
-          "row": 12028
+          "row": 40700
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 4884124
+      "offset": 2048,
+      "documents": 4911305
     },
     "ultra_style": {
-      "page": 1820,
-      "row": 334,
+      "page": 1830,
+      "row": 482,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 9554
+          "row": 50514
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 7455054
+      "documents": 7496162
     },
     "cortex": {
-      "page": 234323,
-      "row": 46,
+      "page": 235633,
+      "row": 81,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 29993390
+      "documents": 30161105
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 895000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "2d68de6c3624467095a90ec9bdeef51e2fc5ab4e71e38707f6c6cc942ef1b8c8", "coverage.json": "da94cd20bec68c816560b61fce43b1062241b6873d42a2e73f505b9d916b5a6b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000895000"}
+{"step": 900000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9", "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000900000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd
+oid sha256:ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7
 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": 76601.26916859495,
-      "required_pages_per_second_upper_estimate": 0.005299453540515585
+      "required_token_positions_per_second": 77365.33532253442,
+      "required_pages_per_second_upper_estimate": 0.005352313409922849
     },
     {
       "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": 9575.15864607437,
-      "required_pages_per_second_upper_estimate": 0.004008918973061503
+      "required_token_positions_per_second": 9670.666915316802,
+      "required_pages_per_second_upper_estimate": 0.004048906290953853
     },
     {
       "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": 19150.31729214874,
-      "required_pages_per_second_upper_estimate": 0.003988506064125937
+      "required_token_positions_per_second": 19341.333830633605,
+      "required_pages_per_second_upper_estimate": 0.004028289771647462
     },
     {
       "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": 28725.47593822311,
-      "required_pages_per_second_upper_estimate": 0.0017220484212343672
+      "required_token_positions_per_second": 29012.00074595041,
+      "required_pages_per_second_upper_estimate": 0.0017392251459595694
     },
     {
       "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": 28725.47593822311,
-      "required_pages_per_second_upper_estimate": 0.0022332794041812756
+      "required_token_positions_per_second": 29012.00074595041,
+      "required_pages_per_second_upper_estimate": 0.002255555447692636
     },
     {
       "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": 28725.47593822311,
-      "required_pages_per_second_upper_estimate": 0.36845956232248317
+      "required_token_positions_per_second": 29012.00074595041,
+      "required_pages_per_second_upper_estimate": 0.3721347949096395
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.09046511610937102,
-    "producer_p95_seconds": 0.19785067599877948,
-    "gpu_update_seconds": 0.9624905801198795,
-    "producer_batches_per_second": 11.053984596571063,
-    "gpu_batches_per_second": 1.0389712072563335,
-    "producer_headroom": 10.639356046989894,
+    "producer_batch_seconds": 0.10793024863289702,
+    "producer_p95_seconds": 0.2442270199990162,
+    "gpu_update_seconds": 0.9529849472328861,
+    "producer_batches_per_second": 9.265243179428767,
+    "gpu_batches_per_second": 1.0493345177210072,
+    "producer_headroom": 8.82963728244778,
     "sustainable": true,
-    "buffer_seconds": 60.636906547552414,
+    "buffer_seconds": 60.038051675671824,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 539194.1073240534,
-    "upload_bytes_per_second": 5396381.570381515,
+    "generation_bytes_per_second": 398086.69224948075,
+    "upload_bytes_per_second": 5232468.799184691,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 115.6985036353226,
+    "estimated_full_shards_per_hour": 85.42013717282597,
     "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": 1.0580149020001954,
-      "bytes": 6239919,
+      "seconds": 1.1932894399997167,
+      "bytes": 6531618,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9808759570000802,
-      "bytes": 5815079,
-      "pages": 28,
+      "seconds": 1.1864704220006388,
+      "bytes": 6136727,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.6121151610000197,
-      "bytes": 12779121,
-      "pages": 35,
-      "shards": 4
+      "seconds": 1.2697396670000671,
+      "bytes": 6182466,
+      "pages": 33,
+      "shards": 3
     },
     {
-      "seconds": 1.091388643999835,
-      "bytes": 6015354,
-      "pages": 36,
+      "seconds": 1.2363136280000617,
+      "bytes": 6022734,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.609872815000017,
-      "bytes": 7812095,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9189778770000885,
-      "bytes": 1227062,
-      "pages": 33,
+      "seconds": 1.2937560119999034,
+      "bytes": 947666,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.2558460740001465,
-      "bytes": 9834347,
-      "pages": 36,
-      "shards": 3
+      "seconds": 1.5761228700002903,
+      "bytes": 18281686,
+      "pages": 39,
+      "shards": 5
     },
     {
-      "seconds": 1.1021917439998106,
-      "bytes": 2521958,
-      "pages": 32,
+      "seconds": 1.1403170100002171,
+      "bytes": 5965661,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.2651067010001498,
-      "bytes": 11919197,
-      "pages": 34,
-      "shards": 3
+      "seconds": 1.088612870000361,
+      "bytes": 2479743,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.3387522560001344,
-      "bytes": 10892645,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.3296366939994186,
+      "bytes": 6147791,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3189305679998142,
-      "bytes": 2654594,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.1267749469989212,
+      "bytes": 894977,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.483484325000063,
-      "bytes": 6080107,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.4399002789996302,
+      "bytes": 10842331,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.1274267459998555,
-      "bytes": 2324297,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.632049243999063,
+      "bytes": 11967346,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.315684739000062,
-      "bytes": 5833727,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.9512228559997311,
+      "bytes": 901267,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.434490552999705,
-      "bytes": 16176474,
-      "pages": 43,
+      "seconds": 1.6983987419989717,
+      "bytes": 12559131,
+      "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 1.33469606700055,
-      "bytes": 2416191,
-      "pages": 31,
+      "seconds": 1.1707839329992566,
+      "bytes": 6060902,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.1527810859997771,
-      "bytes": 6214716,
-      "pages": 33,
+      "seconds": 1.1801391750013863,
+      "bytes": 2749619,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1686728980002954,
+      "bytes": 6121581,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.56234888400013,
-      "bytes": 11134116,
-      "pages": 40,
+      "seconds": 1.3895534080002108,
+      "bytes": 15955272,
+      "pages": 35,
       "shards": 4
     },
     {
-      "seconds": 1.07002194100005,
-      "bytes": 1145768,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.9986490080009389,
+      "bytes": 2653371,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.216849121999985,
-      "bytes": 5739383,
+      "seconds": 1.1648389490001136,
+      "bytes": 5951393,
       "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.4537456819998624,
-      "bytes": 13399952,
-      "pages": 41,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1661972729998524,
-      "bytes": 5666265,
-      "pages": 35,
+      "seconds": 1.1927761070000997,
+      "bytes": 2669572,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.290868875000342,
-      "bytes": 7144629,
-      "pages": 30,
+      "seconds": 1.1629383399995277,
+      "bytes": 10339749,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.0513503590000255,
-      "bytes": 1139099,
-      "pages": 34,
+      "seconds": 1.6129309269999794,
+      "bytes": 5876330,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.996597190999637,
-      "bytes": 6046442,
-      "pages": 28,
+      "seconds": 1.7068213439997635,
+      "bytes": 8291481,
+      "pages": 35,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2685745549988496,
+      "bytes": 6220542,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.1579751299996133,
-      "bytes": 6933490,
+      "seconds": 1.144102094000118,
+      "bytes": 5891615,
       "pages": 30,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.4397287989995675,
-      "bytes": 11874214,
-      "pages": 32,
+      "seconds": 1.2515704560009908,
+      "bytes": 7512167,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0771185140001762,
-      "bytes": 6044727,
-      "pages": 37,
+      "seconds": 1.4014940459983336,
+      "bytes": 5931776,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.0941710249999232,
-      "bytes": 2736868,
-      "pages": 37,
+      "seconds": 1.1321277869992628,
+      "bytes": 2736110,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.062443419000374,
-      "bytes": 5796244,
-      "pages": 20,
-      "shards": 2
+      "seconds": 1.411381372001415,
+      "bytes": 15991084,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.0528361209999275,
-      "bytes": 5176693,
-      "pages": 44,
-      "shards": 2
+      "seconds": 0.8898140989986132,
+      "bytes": 1156490,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 1.123353902999952,
-      "bytes": 7475982,
-      "pages": 39,
+      "seconds": 1.3950513529998716,
+      "bytes": 7312918,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.3653379679999489,
-      "bytes": 11617938,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.2647196729994903,
+      "bytes": 6433157,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.2881032879995473,
-      "bytes": 6772198,
-      "pages": 31,
+      "seconds": 1.2063595479994547,
+      "bytes": 6907543,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.1199771049996343,
-      "bytes": 6074153,
+      "seconds": 1.290804393000144,
+      "bytes": 6092692,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1042831229997319,
-      "bytes": 6145766,
+      "seconds": 1.4008768420007982,
+      "bytes": 11876531,
       "pages": 36,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 1.2581367550001232,
-      "bytes": 2542671,
-      "pages": 33,
+      "seconds": 0.9798814419991686,
+      "bytes": 2581068,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.7602326240003094,
-      "bytes": 10170709,
-      "pages": 35,
+      "seconds": 1.9553985610000382,
+      "bytes": 10094743,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.23988030000055,
-      "bytes": 8293432,
+      "seconds": 1.0605444640004862,
+      "bytes": 2480385,
       "pages": 33,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.1932894399997167,
-      "bytes": 6531618,
-      "pages": 36,
+      "seconds": 1.4249819679989741,
+      "bytes": 6270908,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.1864704220006388,
-      "bytes": 6136727,
-      "pages": 35,
+      "seconds": 1.2829982630009908,
+      "bytes": 5992379,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2697396670000671,
-      "bytes": 6182466,
-      "pages": 33,
-      "shards": 3
+      "seconds": 1.3964128229999915,
+      "bytes": 12688722,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 1.2363136280000617,
-      "bytes": 6022734,
+      "seconds": 1.4358708510008,
+      "bytes": 6164418,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2937560119999034,
-      "bytes": 947666,
-      "pages": 30,
+      "seconds": 1.5897931969993806,
+      "bytes": 7707081,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0018968719996337,
+      "bytes": 927002,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.5761228700002903,
-      "bytes": 18281686,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.3852189129993349,
+      "bytes": 10288556,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.1403170100002171,
-      "bytes": 5965661,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.2743636189989047,
+      "bytes": 2631065,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.088612870000361,
-      "bytes": 2479743,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.369053440999778,
+      "bytes": 11976832,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.3296366939994186,
-      "bytes": 6147791,
-      "pages": 36,
+      "seconds": 1.3680221860013262,
+      "bytes": 5930978,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1267749469989212,
-      "bytes": 894977,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.3103218420001213,
+      "bytes": 6746870,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.4399002789996302,
-      "bytes": 10842331,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.4409255560003658,
+      "bytes": 6062763,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.632049243999063,
-      "bytes": 11967346,
-      "pages": 40,
+      "seconds": 1.1668520560015168,
+      "bytes": 7696896,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9512228559997311,
-      "bytes": 901267,
-      "pages": 26,
+      "seconds": 1.1897505889992317,
+      "bytes": 1090330,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.6983987419989717,
-      "bytes": 12559131,
-      "pages": 36,
+      "seconds": 1.3477415189991007,
+      "bytes": 15713674,
+      "pages": 35,
       "shards": 4
     },
     {
-      "seconds": 1.1707839329992566,
-      "bytes": 6060902,
-      "pages": 32,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1801391750013863,
-      "bytes": 2749619,
-      "pages": 37,
+      "seconds": 1.2074203040010616,
+      "bytes": 7463350,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1686728980002954,
-      "bytes": 6121581,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.190478869000799,
+      "bytes": 1226986,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.3895534080002108,
-      "bytes": 15955272,
-      "pages": 35,
+      "seconds": 1.5480925499996374,
+      "bytes": 11721285,
+      "pages": 32,
       "shards": 4
     },
     {
-      "seconds": 0.9986490080009389,
-      "bytes": 2653371,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.0147747470000468,
+      "bytes": 1162064,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.1648389490001136,
-      "bytes": 5951393,
-      "pages": 27,
+      "seconds": 1.132629542000359,
+      "bytes": 6067377,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1927761070000997,
-      "bytes": 2669572,
+      "seconds": 1.3739411900005507,
+      "bytes": 13347857,
       "pages": 36,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.1629383399995277,
-      "bytes": 10339749,
-      "pages": 40,
-      "shards": 3
+      "seconds": 1.5033581300012884,
+      "bytes": 979335,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 1.6129309269999794,
-      "bytes": 5876330,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.363758608000353,
+      "bytes": 11974676,
+      "pages": 44,
+      "shards": 4
     },
     {
-      "seconds": 1.7068213439997635,
-      "bytes": 8291481,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.1600040840003203,
+      "bytes": 6119662,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.2685745549988496,
-      "bytes": 6220542,
-      "pages": 34,
-      "shards": 2
+      "seconds": 0.8502458040002239,
+      "bytes": 891680,
+      "pages": 27,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985
-size 721774207
+oid sha256:a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d
+size 721766271