Qwen90M step 395000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 390000,
+  "step": 395000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce",
+    "model.safetensors": "90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f",
+    "trainer.pt": "44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730",
     "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
     "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
-    "shard_scheduler.json": "a3cecf4f99b4babbedb1f1e008b99c2bf6e52201ec4ee5a9670c2a4034e567de",
-    "coverage.json": "eb0c2309e40705779a35e534d2ee49b824e6be0073c4849921791384d1631876",
+    "shard_scheduler.json": "236c080251e2cd0c6338dee63d27f4a3f72745e2b8953e3c3cd894d9a6eaa61a",
+    "coverage.json": "e9b422c27949b4fcbed30732a16dcbd7fb3ac93a942ade09b9e407907eebd96b",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 390000,
+  "step": 395000,
   "stats": {
-    "climbmix/blocks": 14040000,
-    "climbmix/targets": 23469768992,
-    "climbmix/input_tokens": 23471255758,
-    "ultra_qa/blocks": 5265000,
-    "ultra_qa/targets": 8589103233,
-    "ultra_qa/input_tokens": 8589119791,
-    "cortex/blocks": 5265000,
-    "cortex/targets": 200266960,
-    "cortex/input_tokens": 8369485865,
-    "ultra_style/blocks": 5265000,
-    "ultra_style/targets": 8328115889,
-    "ultra_style/input_tokens": 8328552449,
-    "openmath/blocks": 3510000,
-    "openmath/targets": 5423887511,
-    "openmath/input_tokens": 5425694857,
-    "rewrite6/blocks": 1755000,
-    "rewrite6/targets": 2877105386,
-    "rewrite6/input_tokens": 2880259808
+    "climbmix/blocks": 14220000,
+    "climbmix/targets": 23770575825,
+    "climbmix/input_tokens": 23772081480,
+    "ultra_qa/blocks": 5332500,
+    "ultra_qa/targets": 8699064740,
+    "ultra_qa/input_tokens": 8699081510,
+    "cortex/blocks": 5332500,
+    "cortex/targets": 202831075,
+    "cortex/input_tokens": 8476791839,
+    "ultra_style/blocks": 5332500,
+    "ultra_style/targets": 8434525911,
+    "ultra_style/input_tokens": 8434968028,
+    "openmath/blocks": 3555000,
+    "openmath/targets": 5493545787,
+    "openmath/input_tokens": 5495376342,
+    "rewrite6/blocks": 1777500,
+    "rewrite6/targets": 2913987031,
+    "rewrite6/input_tokens": 2917181699
   },
   "sources": {
     "climbmix": {
-      "page": 1922,
-      "row": 3973,
+      "page": 1947,
+      "row": 2219,
       "start": {
-        "file": 93,
+        "file": 94,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 3,
+          "row_group": 20,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 6144,
-      "documents": 7876485
+      "offset": 12288,
+      "documents": 7977131
     },
     "rewrite6": {
-      "page": 44,
-      "row": 214,
+      "page": 7,
+      "row": 3678,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 180224
+          "row": 28672
         }
       },
-      "epoch": 21,
+      "epoch": 22,
       "offset": 0,
-      "documents": 4493922
+      "documents": 4551238
     },
     "openmath": {
-      "page": 1201,
-      "row": 3048,
+      "page": 1217,
+      "row": 561,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 116,
-          "row": 146
+          "row_group": 181,
+          "row": 682
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 4922344
+      "documents": 4985393
     },
     "ultra_qa": {
-      "page": 519,
-      "row": 2369,
+      "page": 526,
+      "row": 988,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 0,
-          "row": 47323
+          "row_group": 1,
+          "row": 11508
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 2128193
+      "documents": 2155484
     },
     "ultra_style": {
-      "page": 793,
-      "row": 45,
+      "page": 803,
+      "row": 968,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 82015
+          "row_group": 6,
+          "row": 39519
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 3248173
+      "offset": 2048,
+      "documents": 3290056
     },
     "cortex": {
-      "page": 102116,
-      "row": 47,
+      "page": 103426,
+      "row": 6,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 13070895
+      "documents": 13238534
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 390000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "a3cecf4f99b4babbedb1f1e008b99c2bf6e52201ec4ee5a9670c2a4034e567de", "coverage.json": "eb0c2309e40705779a35e534d2ee49b824e6be0073c4849921791384d1631876", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000390000"}
+{"step": 395000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "236c080251e2cd0c6338dee63d27f4a3f72745e2b8953e3c3cd894d9a6eaa61a", "coverage.json": "e9b422c27949b4fcbed30732a16dcbd7fb3ac93a942ade09b9e407907eebd96b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000395000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce
+oid sha256:90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 75353.21624803833,
-      "required_pages_per_second_upper_estimate": 0.006531425803399414
+      "required_token_positions_per_second": 75571.83458993434,
+      "required_pages_per_second_upper_estimate": 0.0065503750871918395
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9419.15203100479,
-      "required_pages_per_second_upper_estimate": 0.003757440574040526
+      "required_token_positions_per_second": 9446.479323741793,
+      "required_pages_per_second_upper_estimate": 0.0037683418396927525
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18838.30406200958,
-      "required_pages_per_second_upper_estimate": 0.004357910101918019
+      "required_token_positions_per_second": 18892.958647483585,
+      "required_pages_per_second_upper_estimate": 0.004370553478379578
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28257.456093014374,
-      "required_pages_per_second_upper_estimate": 0.0017207308206502374
+      "required_token_positions_per_second": 28339.437971225376,
+      "required_pages_per_second_upper_estimate": 0.0017257230869076143
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28257.456093014374,
-      "required_pages_per_second_upper_estimate": 0.00217533719168512
+      "required_token_positions_per_second": 28339.437971225376,
+      "required_pages_per_second_upper_estimate": 0.0021816483836101666
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28257.456093014374,
-      "required_pages_per_second_upper_estimate": 0.35086303305330935
+      "required_token_positions_per_second": 28339.437971225376,
+      "required_pages_per_second_upper_estimate": 0.35188097360499365
     }
   ],
   "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": 717798,
-    "producer_batch_seconds": 0.09481285885919988,
-    "producer_p95_seconds": 0.3795729499979643,
-    "gpu_update_seconds": 0.9784320254799922,
-    "producer_batches_per_second": 10.547092578286579,
-    "gpu_batches_per_second": 1.022043406142013,
-    "producer_headroom": 10.319613154297931,
+    "producer_batch_seconds": 0.44650785668744675,
+    "producer_p95_seconds": 0.43059379299666034,
+    "gpu_update_seconds": 0.9756015637315238,
+    "producer_batches_per_second": 2.2396022489252525,
+    "gpu_batches_per_second": 1.0250086071768438,
+    "producer_headroom": 2.1849594561881136,
     "sustainable": true,
-    "buffer_seconds": 60.662785579759515,
+    "buffer_seconds": 60.48729695135448,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 285047.90833474684,
-    "upload_bytes_per_second": 5287078.859485369,
+    "generation_bytes_per_second": 483988.12176279345,
+    "upload_bytes_per_second": 5716853.5664507225,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 61.164645552938495,
+    "estimated_full_shards_per_hour": 103.85258426344731,
     "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": 1.0007863089995226,
-      "bytes": 6027863,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.4056276539995451,
+      "bytes": 7428876,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 0.9187128839985235,
-      "bytes": 2707013,
-      "pages": 37,
+      "seconds": 1.1535106260016619,
+      "bytes": 12418046,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 0.8009545050008455,
+      "bytes": 2263071,
+      "pages": 25,
       "shards": 2
     },
     {
-      "seconds": 0.9819577199996274,
-      "bytes": 10903211,
-      "pages": 33,
+      "seconds": 2.506807755002228,
+      "bytes": 10223439,
+      "pages": 45,
       "shards": 3
     },
     {
-      "seconds": 0.860417497002345,
-      "bytes": 864723,
+      "seconds": 0.890203859002213,
+      "bytes": 916282,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.116861483998946,
-      "bytes": 7838954,
+      "seconds": 1.0841833719969145,
+      "bytes": 7541614,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1656662089990277,
-      "bytes": 11880694,
-      "pages": 33,
-      "shards": 3
+      "seconds": 0.8043565109983319,
+      "bytes": 540912,
+      "pages": 18,
+      "shards": 1
     },
     {
-      "seconds": 1.2683707119977043,
-      "bytes": 2325676,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.4004500939990976,
+      "bytes": 16241287,
+      "pages": 47,
+      "shards": 4
     },
     {
-      "seconds": 1.4917878320011368,
-      "bytes": 10275338,
-      "pages": 40,
+      "seconds": 1.0899500639934558,
+      "bytes": 7493063,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.1128048520004086,
-      "bytes": 6259334,
+      "seconds": 0.8779368290051934,
+      "bytes": 1248675,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2020556749994284,
+      "bytes": 12102958,
       "pages": 36,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.2836710900010075,
-      "bytes": 2351070,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.048562693002168,
+      "bytes": 883869,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.2001916299996083,
-      "bytes": 10729681,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.072776924003847,
+      "bytes": 6196482,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0578519230002712,
-      "bytes": 8294494,
-      "pages": 34,
+      "seconds": 0.9420911959969089,
+      "bytes": 7110283,
+      "pages": 25,
       "shards": 3
     },
     {
-      "seconds": 1.1691894300020067,
-      "bytes": 6431968,
-      "pages": 36,
+      "seconds": 1.256377692996466,
+      "bytes": 6950847,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 0.9079114560008748,
-      "bytes": 1141999,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.1828810689985403,
-      "bytes": 7222759,
-      "pages": 29,
+      "seconds": 1.3214142249998986,
+      "bytes": 10934190,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.2303334819989686,
-      "bytes": 10801054,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.2156976079932065,
+      "bytes": 2375462,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 0.9937779290012259,
-      "bytes": 2662341,
+      "seconds": 0.9770713670004625,
+      "bytes": 6199366,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1579458060004981,
-      "bytes": 11572043,
-      "pages": 34,
+      "seconds": 1.0861811289942125,
+      "bytes": 7695547,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8931559880002169,
-      "bytes": 949478,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.0140557359991362,
+      "bytes": 5030935,
+      "pages": 25,
+      "shards": 3
     },
     {
-      "seconds": 1.1832785620026698,
-      "bytes": 12385340,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.2526760300024762,
+      "bytes": 11691054,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.0688240559975384,
-      "bytes": 6186249,
+      "seconds": 0.9398190409992822,
+      "bytes": 2819009,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.054510723999556,
-      "bytes": 1093266,
-      "pages": 33,
-      "shards": 1
+      "seconds": 1.1499773260002257,
+      "bytes": 6017503,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.209129931001371,
-      "bytes": 7450282,
-      "pages": 31,
+      "seconds": 1.6606815510021988,
+      "bytes": 10233181,
+      "pages": 23,
       "shards": 3
     },
     {
-      "seconds": 2.1295198230000096,
-      "bytes": 324019,
+      "seconds": 1.0018192189963884,
+      "bytes": 366522,
       "pages": 10,
       "shards": 1
     },
     {
-      "seconds": 1.2268210219990578,
-      "bytes": 15731128,
-      "pages": 39,
-      "shards": 4
+      "seconds": 0.8059965610009385,
+      "bytes": 2452391,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 0.8699136540017207,
-      "bytes": 2665528,
-      "pages": 36,
+      "seconds": 0.9446101400026237,
+      "bytes": 6258704,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.3528436140004487,
-      "bytes": 6109175,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.1053826549978112,
+      "bytes": 8176480,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.3453164509992348,
-      "bytes": 11716150,
-      "pages": 31,
-      "shards": 4
+      "seconds": 1.249701370994444,
+      "bytes": 10043413,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 0.8494216189974395,
-      "bytes": 1181935,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.0213671140008955,
+      "bytes": 6189327,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 3.5161270839998906,
-      "bytes": 6159420,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.0084248750063125,
+      "bytes": 2332449,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.2511098670001957,
-      "bytes": 8306730,
-      "pages": 34,
-      "shards": 3
+      "seconds": 0.9006169920030516,
+      "bytes": 6118351,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.168453648002469,
-      "bytes": 10300928,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.270539192002616,
+      "bytes": 5502511,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 4.622547731996747,
-      "bytes": 6328963,
+      "seconds": 1.512666299000557,
+      "bytes": 13201899,
       "pages": 33,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 0.9812373249988013,
-      "bytes": 2474379,
-      "pages": 32,
+      "seconds": 1.000070736001362,
+      "bytes": 6089030,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.035560734999308,
-      "bytes": 6144689,
+      "seconds": 1.0753602199984016,
+      "bytes": 871659,
+      "pages": 26,
+      "shards": 1
+    },
+    {
+      "seconds": 1.484286307997536,
+      "bytes": 11527011,
       "pages": 36,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.3160131810000166,
-      "bytes": 6821427,
-      "pages": 32,
-      "shards": 3
+      "seconds": 0.9214101749967085,
+      "bytes": 1181839,
+      "pages": 33,
+      "shards": 1
     },
     {
-      "seconds": 1.3006846259995655,
-      "bytes": 12006518,
-      "pages": 37,
+      "seconds": 1.1708070859967847,
+      "bytes": 7634853,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.8710538570012432,
-      "bytes": 5900566,
-      "pages": 28,
+      "seconds": 1.219036098998913,
+      "bytes": 6506587,
+      "pages": 23,
       "shards": 2
     },
     {
-      "seconds": 1.0173681279993616,
-      "bytes": 2612410,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.234236194002733,
+      "bytes": 11430289,
+      "pages": 42,
+      "shards": 3
     },
     {
-      "seconds": 1.2024259139980131,
-      "bytes": 11156038,
-      "pages": 33,
+      "seconds": 1.0073759259976214,
+      "bytes": 7448462,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 0.9720597020022979,
-      "bytes": 1226766,
-      "pages": 34,
+      "seconds": 0.7627367220047745,
+      "bytes": 1172696,
+      "pages": 36,
       "shards": 1
     },
     {
-      "seconds": 1.4056276539995451,
-      "bytes": 7428876,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.1723295699994196,
+      "bytes": 5714080,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.1535106260016619,
-      "bytes": 12418046,
+      "seconds": 1.1845541919974494,
+      "bytes": 7669056,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8009545050008455,
-      "bytes": 2263071,
-      "pages": 25,
+      "seconds": 0.9817348009964917,
+      "bytes": 4526983,
+      "pages": 22,
       "shards": 2
     },
     {
-      "seconds": 2.506807755002228,
-      "bytes": 10223439,
-      "pages": 45,
-      "shards": 3
+      "seconds": 1.2601451889931923,
+      "bytes": 13743089,
+      "pages": 49,
+      "shards": 4
     },
     {
-      "seconds": 0.890203859002213,
-      "bytes": 916282,
+      "seconds": 0.8212157870002557,
+      "bytes": 890133,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.0841833719969145,
-      "bytes": 7541614,
+      "seconds": 1.1462097159965197,
+      "bytes": 10417894,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1751894160042866,
+      "bytes": 7506454,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8043565109983319,
-      "bytes": 540912,
-      "pages": 18,
+      "seconds": 1.0263809830066748,
+      "bytes": 924946,
+      "pages": 27,
       "shards": 1
     },
     {
-      "seconds": 1.4004500939990976,
-      "bytes": 16241287,
-      "pages": 47,
+      "seconds": 1.1823451020027278,
+      "bytes": 13453015,
+      "pages": 42,
       "shards": 4
     },
     {
-      "seconds": 1.0899500639934558,
-      "bytes": 7493063,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8779368290051934,
-      "bytes": 1248675,
+      "seconds": 1.2355715770027018,
+      "bytes": 9750066,
       "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2020556749994284,
-      "bytes": 12102958,
-      "pages": 36,
       "shards": 4
     },
     {
-      "seconds": 1.048562693002168,
-      "bytes": 883869,
+      "seconds": 1.382024664999335,
+      "bytes": 840789,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.072776924003847,
-      "bytes": 6196482,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9420911959969089,
-      "bytes": 7110283,
-      "pages": 25,
+      "seconds": 1.7087262210043264,
+      "bytes": 7732105,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.256377692996466,
-      "bytes": 6950847,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.1756209270024556,
+      "bytes": 879254,
+      "pages": 27,
+      "shards": 1
     },
     {
-      "seconds": 1.3214142249998986,
-      "bytes": 10934190,
+      "seconds": 1.4754508929981967,
+      "bytes": 10625968,
       "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.2156976079932065,
-      "bytes": 2375462,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9770713670004625,
-      "bytes": 6199366,
-      "pages": 37,
+      "seconds": 0.9398709419983788,
+      "bytes": 2450871,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.0861811289942125,
-      "bytes": 7695547,
+      "seconds": 1.1338124570029322,
+      "bytes": 11954791,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0140557359991362,
-      "bytes": 5030935,
-      "pages": 25,
+      "seconds": 1.1746257199993124,
+      "bytes": 7528355,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2526760300024762,
-      "bytes": 11691054,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.2720379320016946,
+      "bytes": 4790673,
+      "pages": 27,
+      "shards": 2
     },
     {
-      "seconds": 0.9398190409992822,
-      "bytes": 2819009,
+      "seconds": 0.9963719520019367,
+      "bytes": 6449131,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1499773260002257,
-      "bytes": 6017503,
-      "pages": 37,
+      "seconds": 0.9008199800009606,
+      "bytes": 2360873,
+      "pages": 29,
       "shards": 2
+    },
+    {
+      "seconds": 1.193771635000303,
+      "bytes": 12283347,
+      "pages": 43,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f
-size 721770367
+oid sha256:44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730
+size 721765631