Qwen90M step 480000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 475000,
+  "step": 480000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63",
+    "model.safetensors": "e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368",
+    "trainer.pt": "db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d",
     "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
     "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
-    "shard_scheduler.json": "c7547122b453ecc191b4f5a85a0fcc7e7fd9861a4ac8d453704e064cb85f7f50",
-    "coverage.json": "e1b19a902793bcbca1bc3b1a6316d7ddff0a810806697b5e85713b565d8e41a4",
+    "shard_scheduler.json": "99dff589984cd9ecd0f8113aa15c684b2d16a8376d52ae86de45ab20d0a7c84e",
+    "coverage.json": "15b555705988c6d4ac287412f0fe38ff9364ae15871a90a339f590525bb34a89",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 475000,
+  "step": 480000,
   "stats": {
-    "climbmix/blocks": 17100000,
-    "climbmix/targets": 28583228062,
-    "climbmix/input_tokens": 28585039176,
-    "ultra_qa/blocks": 6412500,
-    "ultra_qa/targets": 10460856297,
-    "ultra_qa/input_tokens": 10460876436,
-    "cortex/blocks": 6412500,
-    "cortex/targets": 243889844,
-    "cortex/input_tokens": 10193624611,
-    "ultra_style/blocks": 6412500,
-    "ultra_style/targets": 10143291330,
-    "ultra_style/input_tokens": 10143822186,
-    "openmath/blocks": 4275000,
-    "openmath/targets": 6605793668,
-    "openmath/input_tokens": 6607994980,
-    "rewrite6/blocks": 2137500,
-    "rewrite6/targets": 3504058776,
-    "rewrite6/input_tokens": 3507900626
+    "climbmix/blocks": 17280000,
+    "climbmix/targets": 28884017959,
+    "climbmix/input_tokens": 28885848126,
+    "ultra_qa/blocks": 6480000,
+    "ultra_qa/targets": 10570973360,
+    "ultra_qa/input_tokens": 10570993707,
+    "cortex/blocks": 6480000,
+    "cortex/targets": 246461343,
+    "cortex/input_tokens": 10301013526,
+    "ultra_style/blocks": 6480000,
+    "ultra_style/targets": 10249752091,
+    "ultra_style/input_tokens": 10250288878,
+    "openmath/blocks": 4320000,
+    "openmath/targets": 6675226184,
+    "openmath/input_tokens": 6677450412,
+    "rewrite6/blocks": 2160000,
+    "rewrite6/targets": 3541074887,
+    "rewrite6/input_tokens": 3544957180
   },
   "sources": {
     "climbmix": {
-      "page": 2342,
-      "row": 1120,
+      "page": 2366,
+      "row": 3875,
       "start": {
-        "file": 113,
+        "file": 114,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 28,
+          "row_group": 41,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 9593952
+      "documents": 9695011
     },
     "rewrite6": {
-      "page": 32,
-      "row": 1761,
+      "page": 46,
+      "row": 1940,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 131072
+          "row": 188416
         }
       },
       "epoch": 26,
       "offset": 2048,
-      "documents": 5473337
+      "documents": 5530860
     },
     "openmath": {
-      "page": 1463,
-      "row": 3070,
+      "page": 1479,
+      "row": 475,
       "start": {
         "file": 13,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 315,
-          "row": 998
+          "row_group": 381,
+          "row": 534
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 5995518
+      "documents": 6058459
     },
     "ultra_qa": {
-      "page": 632,
-      "row": 3515,
+      "page": 639,
+      "row": 2127,
       "start": {
-        "file": 4,
+        "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 7,
-          "row": 57974
+          "row_group": 0,
+          "row": 19215
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2592187
+      "offset": 4096,
+      "documents": 2619471
     },
     "ultra_style": {
-      "page": 965,
-      "row": 1470,
+      "page": 975,
+      "row": 2846,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 18773
+          "row": 59733
         }
       },
       "epoch": 0,
       "offset": 4096,
-      "documents": 3954110
+      "documents": 3996446
     },
     "cortex": {
-      "page": 124371,
-      "row": 95,
+      "page": 125681,
+      "row": 34,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 15919583
+      "documents": 16087202
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 475000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "c7547122b453ecc191b4f5a85a0fcc7e7fd9861a4ac8d453704e064cb85f7f50", "coverage.json": "e1b19a902793bcbca1bc3b1a6316d7ddff0a810806697b5e85713b565d8e41a4", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000475000"}
+{"step": 480000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "99dff589984cd9ecd0f8113aa15c684b2d16a8376d52ae86de45ab20d0a7c84e", "coverage.json": "15b555705988c6d4ac287412f0fe38ff9364ae15871a90a339f590525bb34a89", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000480000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63
+oid sha256:e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a
 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": 75488.53687979521,
-      "required_pages_per_second_upper_estimate": 0.005222472138881793
+      "required_token_positions_per_second": 75948.0229487074,
+      "required_pages_per_second_upper_estimate": 0.005254260451283705
     },
     {
       "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": 9436.067109974401,
-      "required_pages_per_second_upper_estimate": 0.003950684251458008
+      "required_token_positions_per_second": 9493.502868588424,
+      "required_pages_per_second_upper_estimate": 0.003974731404194673
     },
     {
       "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": 18872.134219948803,
-      "required_pages_per_second_upper_estimate": 0.003930567866367642
+      "required_token_positions_per_second": 18987.00573717685,
+      "required_pages_per_second_upper_estimate": 0.003954492574040618
     },
     {
       "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": 28308.201329923202,
-      "required_pages_per_second_upper_estimate": 0.001697033445608223
+      "required_token_positions_per_second": 28480.508605765273,
+      "required_pages_per_second_upper_estimate": 0.0017073630036969789
     },
     {
       "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": 28308.201329923202,
-      "required_pages_per_second_upper_estimate": 0.002200838138782989
+      "required_token_positions_per_second": 28480.508605765273,
+      "required_pages_per_second_upper_estimate": 0.002214234271580101
     },
     {
       "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": 28308.201329923202,
-      "required_pages_per_second_upper_estimate": 0.3631072116817794
+      "required_token_positions_per_second": 28480.508605765273,
+      "required_pages_per_second_upper_estimate": 0.36531738440714295
     }
   ],
   "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": 0.1093707058749942,
-    "producer_p95_seconds": 0.32194404499750817,
-    "gpu_update_seconds": 0.9766780897793977,
-    "producer_batches_per_second": 9.143216110746831,
-    "gpu_batches_per_second": 1.0238788096760418,
-    "producer_headroom": 8.92997884548443,
+    "producer_batch_seconds": 0.14875617208599579,
+    "producer_p95_seconds": 0.35381016099563567,
+    "gpu_update_seconds": 0.970769180519594,
+    "producer_batches_per_second": 6.722410142564714,
+    "gpu_batches_per_second": 1.030110988345098,
+    "producer_headroom": 6.525908585214155,
     "sustainable": true,
-    "buffer_seconds": 60.55404156632266,
+    "buffer_seconds": 60.18768919221483,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 85,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 399042.23246710177,
-    "upload_bytes_per_second": 5347906.198548627,
+    "generation_bytes_per_second": 352449.9817351,
+    "upload_bytes_per_second": 5306492.160540054,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 85.62517385968961,
+    "estimated_full_shards_per_hour": 75.62756146468877,
     "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": 1.1180898200000229,
-      "bytes": 7687731,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1014808299987635,
-      "bytes": 5892940,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.3809676959972421,
+      "bytes": 13424902,
+      "pages": 35,
+      "shards": 4
     },
     {
-      "seconds": 0.8753045230005227,
-      "bytes": 5867523,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.9754727400031697,
+      "bytes": 994247,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.8846877780015348,
-      "bytes": 8419016,
-      "pages": 36,
+      "seconds": 1.2377208289981354,
+      "bytes": 11067407,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9963944320006703,
-      "bytes": 6337215,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0649026120008784,
-      "bytes": 9997467,
-      "pages": 31,
+      "seconds": 1.5320398900003056,
+      "bytes": 7616229,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.0817448720008542,
-      "bytes": 2541266,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.994875273998332,
+      "bytes": 1060442,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.4285129730014887,
-      "bytes": 6270013,
+      "seconds": 1.0037741340020148,
+      "bytes": 6334013,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.485799860998668,
-      "bytes": 7350955,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.3855582350006443,
-      "bytes": 11926342,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.4788876900020114,
+      "bytes": 12280922,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.1727449809986865,
-      "bytes": 5685962,
+      "seconds": 1.3189756050014694,
+      "bytes": 5844934,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2225738800007093,
-      "bytes": 2613097,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0333214540005429,
-      "bytes": 6358222,
-      "pages": 38,
-      "shards": 2
+      "seconds": 1.142625862001296,
+      "bytes": 7700249,
+      "pages": 41,
+      "shards": 3
     },
     {
-      "seconds": 1.2548035990002973,
-      "bytes": 11160297,
-      "pages": 33,
-      "shards": 4
+      "seconds": 0.8653922449957463,
+      "bytes": 888116,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.0187819040002069,
-      "bytes": 1294212,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3053653089955333,
+      "bytes": 10336652,
+      "pages": 41,
+      "shards": 3
     },
     {
-      "seconds": 1.2491005060001044,
-      "bytes": 11541986,
-      "pages": 30,
+      "seconds": 1.1407873469943297,
+      "bytes": 7506775,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.6100285119973705,
-      "bytes": 7404435,
-      "pages": 39,
+      "seconds": 1.0540632260017446,
+      "bytes": 6792818,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.136316104002617,
-      "bytes": 5061136,
+      "seconds": 0.9928698779986007,
+      "bytes": 5905666,
       "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2786440780000703,
-      "bytes": 7516874,
-      "pages": 38,
+      "seconds": 2.588622392999241,
+      "bytes": 7185173,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 0.9931387870019535,
-      "bytes": 6287159,
-      "pages": 38,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9191857780024293,
-      "bytes": 859128,
-      "pages": 26,
+      "seconds": 1.1344652970001334,
+      "bytes": 5709671,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.4843631869989622,
-      "bytes": 18084349,
-      "pages": 33,
-      "shards": 5
+      "seconds": 1.2757203309956822,
+      "bytes": 7569977,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 0.9931742519984255,
-      "bytes": 1121111,
+      "seconds": 0.8296898909975425,
+      "bytes": 1129072,
       "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.0368349029995443,
-      "bytes": 5950136,
-      "pages": 29,
-      "shards": 2
+      "seconds": 1.0205191320055746,
+      "bytes": 9233133,
+      "pages": 13,
+      "shards": 3
     },
     {
-      "seconds": 0.8421170019973943,
-      "bytes": 2576016,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.1241545290031354,
+      "bytes": 8856234,
+      "pages": 49,
+      "shards": 3
     },
     {
-      "seconds": 0.9882992809980351,
-      "bytes": 6015320,
-      "pages": 28,
+      "seconds": 1.2785397469997406,
+      "bytes": 5725754,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.0212832630022604,
-      "bytes": 4591085,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.0718349349990604,
+      "bytes": 7504941,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.4545433539969963,
-      "bytes": 12931413,
-      "pages": 25,
-      "shards": 4
+      "seconds": 1.2024066430021776,
+      "bytes": 5964235,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.2126570350010297,
-      "bytes": 6557056,
-      "pages": 45,
+      "seconds": 1.2837438990027294,
+      "bytes": 5816020,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 0.8083091259977664,
-      "bytes": 741591,
-      "pages": 21,
+      "seconds": 1.0380413320017396,
+      "bytes": 232525,
+      "pages": 6,
       "shards": 1
     },
     {
-      "seconds": 1.6113828489978914,
-      "bytes": 12223061,
-      "pages": 49,
-      "shards": 4
+      "seconds": 1.743612624995876,
+      "bytes": 7907888,
+      "pages": 23,
+      "shards": 3
     },
     {
-      "seconds": 0.9428402260018629,
-      "bytes": 5721024,
-      "pages": 29,
+      "seconds": 2.2645728169954964,
+      "bytes": 6235357,
+      "pages": 44,
       "shards": 2
     },
     {
-      "seconds": 1.6995308779987681,
-      "bytes": 2763568,
+      "seconds": 1.3341411860019434,
+      "bytes": 11745696,
       "pages": 39,
-      "shards": 2
+      "shards": 4
+    },
+    {
+      "seconds": 3.1945914269963396,
+      "bytes": 906980,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.0716314189994591,
-      "bytes": 5628326,
-      "pages": 20,
+      "seconds": 1.100945904996479,
+      "bytes": 6366458,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0639040940004634,
-      "bytes": 11047745,
-      "pages": 29,
+      "seconds": 1.1544355079968227,
+      "bytes": 7474922,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.0850444019997667,
-      "bytes": 7634525,
+      "seconds": 1.0820681480036,
+      "bytes": 10987528,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 4.141720985000575,
-      "bytes": 6233393,
+      "seconds": 1.2853161590028321,
+      "bytes": 6397328,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9377798560017254,
-      "bytes": 884750,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.367523049000738,
-      "bytes": 11388768,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.1332088359995396,
+      "bytes": 7397305,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 0.9475808069983032,
-      "bytes": 1235313,
-      "pages": 36,
+      "seconds": 0.7730173130039475,
+      "bytes": 1053057,
+      "pages": 32,
       "shards": 1
     },
     {
-      "seconds": 1.3809676959972421,
-      "bytes": 13424902,
-      "pages": 35,
+      "seconds": 1.3466375830030302,
+      "bytes": 11763121,
+      "pages": 37,
       "shards": 4
     },
     {
-      "seconds": 0.9754727400031697,
-      "bytes": 994247,
-      "pages": 29,
+      "seconds": 1.4802930930018192,
+      "bytes": 1138429,
+      "pages": 32,
       "shards": 1
     },
     {
-      "seconds": 1.2377208289981354,
-      "bytes": 11067407,
-      "pages": 30,
+      "seconds": 1.8897758500024793,
+      "bytes": 11833830,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.5320398900003056,
-      "bytes": 7616229,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.1506007540010614,
+      "bytes": 2440099,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 0.994875273998332,
-      "bytes": 1060442,
-      "pages": 31,
-      "shards": 1
+      "seconds": 1.5953800970019074,
+      "bytes": 10985431,
+      "pages": 43,
+      "shards": 3
     },
     {
-      "seconds": 1.0037741340020148,
-      "bytes": 6334013,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.5438401100036572,
+      "bytes": 7453011,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.4788876900020114,
-      "bytes": 12280922,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.0392796939995605,
+      "bytes": 1282728,
+      "pages": 37,
+      "shards": 1
     },
     {
-      "seconds": 1.3189756050014694,
-      "bytes": 5844934,
-      "pages": 28,
+      "seconds": 1.3358671570022125,
+      "bytes": 5881131,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.142625862001296,
-      "bytes": 7700249,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.5660001680007554,
+      "bytes": 16980575,
+      "pages": 44,
+      "shards": 6
     },
     {
-      "seconds": 0.8653922449957463,
-      "bytes": 888116,
+      "seconds": 0.8579309720007586,
+      "bytes": 927462,
       "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.3053653089955333,
-      "bytes": 10336652,
-      "pages": 41,
+      "seconds": 1.3555858160034404,
+      "bytes": 7535281,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1407873469943297,
-      "bytes": 7506775,
+      "seconds": 1.1513379759999225,
+      "bytes": 10862357,
       "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0540632260017446,
-      "bytes": 6792818,
-      "pages": 30,
+      "seconds": 0.9964048779947916,
+      "bytes": 1334873,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2506730089953635,
+      "bytes": 7336892,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 0.9928698779986007,
-      "bytes": 5905666,
-      "pages": 30,
+      "seconds": 1.3737510540013318,
+      "bytes": 6860686,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 2.588622392999241,
-      "bytes": 7185173,
+      "seconds": 1.2590301210002508,
+      "bytes": 10720957,
       "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1344652970001334,
-      "bytes": 5709671,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2757203309956822,
-      "bytes": 7569977,
-      "pages": 40,
+      "seconds": 1.187173960999644,
+      "bytes": 7325771,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 0.8296898909975425,
-      "bytes": 1129072,
-      "pages": 35,
+      "seconds": 0.8256457960014814,
+      "bytes": 1213741,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.0205191320055746,
-      "bytes": 9233133,
-      "pages": 13,
+      "seconds": 1.0744473489976372,
+      "bytes": 7318826,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.1241545290031354,
-      "bytes": 8856234,
-      "pages": 49,
+      "seconds": 1.3007696439963183,
+      "bytes": 9895316,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.2785397469997406,
-      "bytes": 5725754,
-      "pages": 29,
+      "seconds": 1.3112097839984926,
+      "bytes": 6821672,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.0718349349990604,
-      "bytes": 7504941,
-      "pages": 39,
+      "seconds": 1.0360545559960883,
+      "bytes": 7415725,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.2024066430021776,
-      "bytes": 5964235,
-      "pages": 32,
+      "seconds": 0.9522623859957093,
+      "bytes": 963026,
+      "pages": 30,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1533947639982216,
+      "bytes": 11916585,
+      "pages": 41,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1312003819984966,
+      "bytes": 5813649,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.2837438990027294,
-      "bytes": 5816020,
+      "seconds": 1.0730638719978742,
+      "bytes": 1187099,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1143769740010612,
+      "bytes": 13033863,
       "pages": 32,
-      "shards": 2
+      "shards": 4
+    },
+    {
+      "seconds": 1.1261608329950832,
+      "bytes": 10613783,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0124226469997666,
+      "bytes": 904800,
+      "pages": 27,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368
-size 721768703
+oid sha256:db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d
+size 721770047