Qwen90M step 665000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 660000,
+  "step": 665000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "72f49cb391b6aab9bca2868aeeef6f213b458317d5990a7777deaf41e238182b",
+    "model.safetensors": "6c241dc1c1a846d32a12afafc492600d69a86e3be4fee8d141a618e8afd44783",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "65de0d9a1fae3b3f97e45bb3d262897c6e998a420cc1f341ac0fc0a061ba545b",
+    "trainer.pt": "717cb301aeb14d38409ae0e0d04856c0c71f8067a7c0afbf8c05d0053df1f880",
     "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
     "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf",
-    "shard_scheduler.json": "20fc220c088b230512fe9811d1be43e07c8d4d163441b00d2987e0cc07f7b87a",
-    "coverage.json": "55c2cb18b6d4418f8d8593a610088ab79f86e62560fed2014b84f1bc8b405a4a",
+    "shard_scheduler.json": "e7d4e34ca9c5186534fb32e800aad10548f8a5c95b53fbc5af10a7ccc7febab0",
+    "coverage.json": "ed08e0edb4aa125a6baee640d15e6403fa05eecc9fd2bba36beb390a00d0f15e",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 660000,
+  "step": 665000,
   "stats": {
-    "climbmix/blocks": 23760000,
-    "climbmix/targets": 39719704239,
-    "climbmix/input_tokens": 39722221283,
-    "ultra_qa/blocks": 8910000,
-    "ultra_qa/targets": 14536285905,
-    "ultra_qa/input_tokens": 14536313968,
-    "cortex/blocks": 8910000,
-    "cortex/targets": 338791948,
-    "cortex/input_tokens": 14163040960,
-    "ultra_style/blocks": 8910000,
-    "ultra_style/targets": 14093005370,
-    "ultra_style/input_tokens": 14093744516,
-    "openmath/blocks": 5940000,
-    "openmath/targets": 9178382826,
-    "openmath/input_tokens": 9181440587,
-    "rewrite6/blocks": 2970000,
-    "rewrite6/targets": 4868912541,
-    "rewrite6/input_tokens": 4874250530
+    "climbmix/blocks": 23940000,
+    "climbmix/targets": 40021124917,
+    "climbmix/input_tokens": 40023660910,
+    "ultra_qa/blocks": 8977500,
+    "ultra_qa/targets": 14646361182,
+    "ultra_qa/input_tokens": 14646389468,
+    "cortex/blocks": 8977500,
+    "cortex/targets": 341355085,
+    "cortex/input_tokens": 14270346565,
+    "ultra_style/blocks": 8977500,
+    "ultra_style/targets": 14200262565,
+    "ultra_style/input_tokens": 14201007196,
+    "openmath/blocks": 5985000,
+    "openmath/targets": 9247885163,
+    "openmath/input_tokens": 9250966078,
+    "rewrite6/blocks": 2992500,
+    "rewrite6/targets": 4905783456,
+    "rewrite6/input_tokens": 4911161965
   },
   "sources": {
     "climbmix": {
-      "page": 3253,
-      "row": 3710,
+      "page": 3278,
+      "row": 1762,
       "start": {
-        "file": 157,
+        "file": 158,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 38,
+          "row_group": 55,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 13327998
+      "offset": 0,
+      "documents": 13428450
     },
     "rewrite6": {
-      "page": 1,
-      "row": 761,
+      "page": 15,
+      "row": 1069,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 4096
+          "row": 61440
         }
       },
       "epoch": 37,
-      "offset": 0,
-      "documents": 7604805
+      "offset": 2048,
+      "documents": 7662457
     },
     "openmath": {
-      "page": 2033,
-      "row": 2387,
+      "page": 2048,
+      "row": 3984,
       "start": {
         "file": 19,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 30,
-          "row": 818
+          "row_group": 92,
+          "row": 258
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 8329555
+      "documents": 8392592
     },
     "ultra_qa": {
-      "page": 879,
-      "row": 1432,
+      "page": 886,
+      "row": 34,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 30603
+          "row": 59275
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 3601816
+      "offset": 2048,
+      "documents": 3629090
     },
     "ultra_style": {
-      "page": 1342,
-      "row": 136,
+      "page": 1352,
+      "row": 176,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 17298
+          "row": 58258
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 5496968
+      "offset": 2048,
+      "documents": 5537968
     },
     "cortex": {
-      "page": 172802,
-      "row": 10,
+      "page": 174111,
+      "row": 108,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 22118666
+      "documents": 22286316
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 660000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "72f49cb391b6aab9bca2868aeeef6f213b458317d5990a7777deaf41e238182b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "65de0d9a1fae3b3f97e45bb3d262897c6e998a420cc1f341ac0fc0a061ba545b", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf", "shard_scheduler.json": "20fc220c088b230512fe9811d1be43e07c8d4d163441b00d2987e0cc07f7b87a", "coverage.json": "55c2cb18b6d4418f8d8593a610088ab79f86e62560fed2014b84f1bc8b405a4a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000660000"}
+{"step": 665000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6c241dc1c1a846d32a12afafc492600d69a86e3be4fee8d141a618e8afd44783", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "717cb301aeb14d38409ae0e0d04856c0c71f8067a7c0afbf8c05d0053df1f880", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "5949473856ba60c6ea6e09958b880f6c16bf1db77e55259c9de9f88123633edf", "shard_scheduler.json": "e7d4e34ca9c5186534fb32e800aad10548f8a5c95b53fbc5af10a7ccc7febab0", "coverage.json": "ed08e0edb4aa125a6baee640d15e6403fa05eecc9fd2bba36beb390a00d0f15e", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000665000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72f49cb391b6aab9bca2868aeeef6f213b458317d5990a7777deaf41e238182b
+oid sha256:6c241dc1c1a846d32a12afafc492600d69a86e3be4fee8d141a618e8afd44783
 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": 76638.2072425708,
-      "required_pages_per_second_upper_estimate": 0.005302009002181374
+      "required_token_positions_per_second": 77402.7761039638,
+      "required_pages_per_second_upper_estimate": 0.005354903650056715
     },
     {
       "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": 9579.77590532135,
-      "required_pages_per_second_upper_estimate": 0.004010852123088876
+      "required_token_positions_per_second": 9675.347012995475,
+      "required_pages_per_second_upper_estimate": 0.00405086575012036
     },
     {
       "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": 19159.5518106427,
-      "required_pages_per_second_upper_estimate": 0.0039904293707976
+      "required_token_positions_per_second": 19350.69402599095,
+      "required_pages_per_second_upper_estimate": 0.004030239253495446
     },
     {
       "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": 28739.32771596405,
-      "required_pages_per_second_upper_estimate": 0.0017228788141594965
+      "required_token_positions_per_second": 29026.041038986423,
+      "required_pages_per_second_upper_estimate": 0.0017400668400887923
     },
     {
       "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": 28739.32771596405,
-      "required_pages_per_second_upper_estimate": 0.0022343563189731025
+      "required_token_positions_per_second": 29026.041038986423,
+      "required_pages_per_second_upper_estimate": 0.002256647018719463
     },
     {
       "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": 28739.32771596405,
-      "required_pages_per_second_upper_estimate": 0.36863723805446375
+      "required_token_positions_per_second": 29026.041038986423,
+      "required_pages_per_second_upper_estimate": 0.3723148887134134
     }
   ],
   "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": 0.06065465613274057,
-    "producer_p95_seconds": 0.20881386700057192,
-    "gpu_update_seconds": 0.9620266790249988,
-    "producer_batches_per_second": 16.48678046762866,
-    "gpu_batches_per_second": 1.0394722119489312,
-    "producer_headroom": 15.860722661087015,
+    "producer_batch_seconds": 0.08651116098442913,
+    "producer_p95_seconds": 0.24094921500000055,
+    "gpu_update_seconds": 0.9525239753800561,
+    "producer_batches_per_second": 11.559202172538026,
+    "gpu_batches_per_second": 1.0498423408198214,
+    "producer_headroom": 11.0104172056077,
     "sustainable": true,
-    "buffer_seconds": 60.607680778574924,
+    "buffer_seconds": 60.009010448943535,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 86,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 517212.2460824716,
-    "upload_bytes_per_second": 2200623.9726313176,
+    "generation_bytes_per_second": 357946.5645164134,
+    "upload_bytes_per_second": 1903783.0719585211,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 110.98170792441951,
+    "estimated_full_shards_per_hour": 76.80700017566014,
     "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": 1.2592496409997693,
-      "bytes": 5842307,
-      "pages": 27,
+      "seconds": 1.164695950999885,
+      "bytes": 11623258,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1422205050002958,
+      "bytes": 2673736,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4229377800002112,
+      "bytes": 5867572,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3798525270003665,
-      "bytes": 2740249,
-      "pages": 37,
+      "seconds": 1.0479602760005946,
+      "bytes": 10355941,
+      "pages": 43,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0539684499999566,
+      "bytes": 2291279,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.102272197000275,
-      "bytes": 5978754,
-      "pages": 29,
+      "seconds": 1.055462077001721,
+      "bytes": 6384191,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.5570587419997537,
-      "bytes": 16077242,
-      "pages": 41,
+      "seconds": 1.3574712270001328,
+      "bytes": 12582265,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 1.4845605279997471,
-      "bytes": 2728370,
+      "seconds": 1.1109151510008815,
+      "bytes": 6070099,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.968915202000062,
-      "bytes": 6356080,
-      "pages": 36,
+      "seconds": 1.0409215150011732,
+      "bytes": 6007655,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.188576070999261,
-      "bytes": 5752447,
-      "pages": 32,
+      "seconds": 0.9865703520008537,
+      "bytes": 2607815,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4639023600002474,
-      "bytes": 7690441,
-      "pages": 33,
+      "seconds": 1.1019334159991558,
+      "bytes": 10289866,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9298561730001893,
-      "bytes": 6431657,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.255959656999039,
-      "bytes": 8183161,
-      "pages": 33,
+      "seconds": 1.2059528549998504,
+      "bytes": 7149120,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.190589429001193,
-      "bytes": 6346216,
-      "pages": 35,
-      "shards": 3
+      "seconds": 0.9931068189998769,
+      "bytes": 6968332,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.0324392470010935,
-      "bytes": 4954714,
+      "seconds": 1.032158955000341,
+      "bytes": 6117925,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.8339393939986621,
-      "bytes": 7623870,
-      "pages": 30,
+      "seconds": 122.56198037699869,
+      "bytes": 6934679,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.1467835090006702,
-      "bytes": 6065504,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.1427723569995578,
+      "bytes": 13775084,
+      "pages": 72,
+      "shards": 3
     },
     {
-      "seconds": 1.5038798329987912,
-      "bytes": 7468080,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.2821505469983094,
+      "bytes": 1156096,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.3404178419987147,
-      "bytes": 11740523,
+      "seconds": 1.3412859470008698,
+      "bytes": 15865632,
       "pages": 31,
+      "shards": 4
+    },
+    {
+      "seconds": 1.324599482000849,
+      "bytes": 7784072,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.198196088000259,
-      "bytes": 6150655,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.8532274739991408,
+      "bytes": 1099982,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.157564569999522,
-      "bytes": 2379174,
+      "seconds": 1.0939723810006399,
+      "bytes": 5792220,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1466768219997903,
-      "bytes": 6100233,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.519392225000047,
+      "bytes": 12323148,
+      "pages": 43,
+      "shards": 4
     },
     {
-      "seconds": 1.4511324699997203,
-      "bytes": 6333083,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.8898182180000731,
+      "bytes": 896140,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.0026024300004792,
-      "bytes": 6038075,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.8745819040013885,
+      "bytes": 13367716,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.4379317339989939,
-      "bytes": 11505212,
-      "pages": 31,
-      "shards": 3
+      "seconds": 0.9403650779986492,
+      "bytes": 1055784,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 1.3395487400011916,
-      "bytes": 6734300,
-      "pages": 31,
+      "seconds": 1.317758003999188,
+      "bytes": 10010526,
+      "pages": 18,
       "shards": 3
     },
     {
-      "seconds": 0.9863095519995113,
-      "bytes": 6175312,
+      "seconds": 1.4487697540007503,
+      "bytes": 2537038,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5478082320005342,
-      "bytes": 6142048,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 0.98161319600149,
-      "bytes": 2080200,
-      "pages": 19,
+      "seconds": 1.5384773589994438,
+      "bytes": 6037444,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.4949951670005248,
-      "bytes": 10484718,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5762320020003244,
-      "bytes": 8331124,
+      "seconds": 0.9018981910012371,
+      "bytes": 1284941,
       "pages": 35,
-      "shards": 3
+      "shards": 1
     },
     {
-      "seconds": 1.3131415360003302,
-      "bytes": 5988822,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.8045908780004538,
+      "bytes": 17039792,
+      "pages": 37,
+      "shards": 5
     },
     {
-      "seconds": 1.0547332340011053,
-      "bytes": 6465464,
-      "pages": 36,
+      "seconds": 1.6887772259997291,
+      "bytes": 6318440,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.862101573999098,
-      "bytes": 2178443,
-      "pages": 22,
-      "shards": 2
+      "seconds": 1.1342329689996404,
+      "bytes": 2581964,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 0.9640832400000363,
-      "bytes": 4456954,
-      "pages": 18,
+      "seconds": 1.048425811000925,
+      "bytes": 5676510,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.0857610709990695,
-      "bytes": 6125796,
-      "pages": 28,
-      "shards": 2
+      "seconds": 33.502104078999764,
+      "bytes": 11160724,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 1.0003182039999956,
-      "bytes": 2721093,
-      "pages": 36,
+      "seconds": 0.9322519389988884,
+      "bytes": 2962222,
+      "pages": 46,
       "shards": 2
     },
     {
-      "seconds": 1.0431637549991137,
-      "bytes": 11688824,
-      "pages": 32,
+      "seconds": 1.3943822889996227,
+      "bytes": 11929043,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.3166948790003516,
-      "bytes": 11002922,
-      "pages": 42,
+      "seconds": 1.140193093000562,
+      "bytes": 7462921,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9981318899990583,
-      "bytes": 2367436,
-      "pages": 28,
+      "seconds": 1.6433775559999049,
+      "bytes": 5520234,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1174767350003094,
-      "bytes": 6113163,
+      "seconds": 1.1214206500008004,
+      "bytes": 6053975,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3073732009997912,
-      "bytes": 7581885,
-      "pages": 39,
+      "seconds": 1.0533692200006044,
+      "bytes": 7678729,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.4157619699999486,
-      "bytes": 4898674,
-      "pages": 29,
-      "shards": 2
+      "seconds": 0.9111081320006633,
+      "bytes": 972942,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.164695950999885,
-      "bytes": 11623258,
+      "seconds": 1.5192261520023749,
+      "bytes": 17166243,
       "pages": 34,
-      "shards": 3
+      "shards": 5
     },
     {
-      "seconds": 1.1422205050002958,
-      "bytes": 2673736,
-      "pages": 36,
-      "shards": 3
+      "seconds": 0.9617387159996724,
+      "bytes": 1097341,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.4229377800002112,
-      "bytes": 5867572,
-      "pages": 28,
+      "seconds": 1.1497930719997385,
+      "bytes": 6229541,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0479602760005946,
-      "bytes": 10355941,
-      "pages": 43,
+      "seconds": 1.2241392690011708,
+      "bytes": 7522866,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0539684499999566,
-      "bytes": 2291279,
-      "pages": 28,
+      "seconds": 1.1316358820004098,
+      "bytes": 5807643,
+      "pages": 41,
       "shards": 2
     },
     {
-      "seconds": 1.055462077001721,
-      "bytes": 6384191,
-      "pages": 37,
+      "seconds": 1.0283135649988253,
+      "bytes": 5734251,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.3574712270001328,
-      "bytes": 12582265,
-      "pages": 37,
+      "seconds": 1.5221587859996362,
+      "bytes": 13580138,
+      "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 1.1109151510008815,
-      "bytes": 6070099,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0409215150011732,
-      "bytes": 6007655,
+      "seconds": 1.401579060999211,
+      "bytes": 925271,
       "pages": 28,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 0.9865703520008537,
-      "bytes": 2607815,
+      "seconds": 1.7116284690018801,
+      "bytes": 11540303,
       "pages": 37,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.1019334159991558,
-      "bytes": 10289866,
-      "pages": 39,
-      "shards": 3
+      "seconds": 0.9012360309970973,
+      "bytes": 1247702,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.2059528549998504,
-      "bytes": 7149120,
-      "pages": 30,
+      "seconds": 1.5333594760013511,
+      "bytes": 7425995,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9931068189998769,
-      "bytes": 6968332,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.032158955000341,
-      "bytes": 6117925,
-      "pages": 37,
+      "seconds": 1.141444616998342,
+      "bytes": 5373163,
+      "pages": 23,
       "shards": 2
     },
     {
-      "seconds": 122.56198037699869,
-      "bytes": 6934679,
-      "pages": 34,
+      "seconds": 1.302073504000873,
+      "bytes": 12612142,
+      "pages": 45,
       "shards": 3
     },
     {
-      "seconds": 1.1427723569995578,
-      "bytes": 13775084,
-      "pages": 72,
+      "seconds": 1.2979291070005274,
+      "bytes": 7799504,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.2821505469983094,
-      "bytes": 1156096,
-      "pages": 35,
+      "seconds": 0.9238773860015499,
+      "bytes": 1050494,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.3412859470008698,
-      "bytes": 15865632,
+      "seconds": 1.0488793139993504,
+      "bytes": 6030424,
       "pages": 31,
-      "shards": 4
+      "shards": 2
     },
     {
-      "seconds": 1.324599482000849,
-      "bytes": 7784072,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.2485899509993033,
+      "bytes": 11733880,
+      "pages": 45,
+      "shards": 4
     },
     {
-      "seconds": 0.8532274739991408,
-      "bytes": 1099982,
-      "pages": 34,
+      "seconds": 1.0325014549998741,
+      "bytes": 941600,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.0939723810006399,
-      "bytes": 5792220,
-      "pages": 28,
+      "seconds": 1.327213976001076,
+      "bytes": 13471794,
+      "pages": 42,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1652102200023364,
+      "bytes": 5705697,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.519392225000047,
-      "bytes": 12323148,
-      "pages": 43,
-      "shards": 4
+      "seconds": 1.1611206180023146,
+      "bytes": 6127791,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 0.8898182180000731,
-      "bytes": 896140,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.3078043249988696,
+      "bytes": 7396085,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.8745819040013885,
-      "bytes": 13367716,
-      "pages": 39,
-      "shards": 4
+      "seconds": 0.9337214369988942,
+      "bytes": 1171128,
+      "pages": 35,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:65de0d9a1fae3b3f97e45bb3d262897c6e998a420cc1f341ac0fc0a061ba545b
-size 721773439
+oid sha256:717cb301aeb14d38409ae0e0d04856c0c71f8067a7c0afbf8c05d0053df1f880
+size 721779967