Qwen90M step 740000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 735000,
+  "step": 740000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73",
+    "model.safetensors": "965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315",
+    "trainer.pt": "7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393",
     "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037",
     "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379",
-    "shard_scheduler.json": "541a1f3f705d2b085cbb51195f8b00accd55d018941e78dc4445045fc4c03dcd",
-    "coverage.json": "fe7d1dc256a120fa7e42273b21198470f2acfeaaf06e57edd8a92aeb71326f7c",
+    "shard_scheduler.json": "7f8712a7f9538f373e624550e1f5a9ce3a8e415d6a140e719f8256eddac3ad48",
+    "coverage.json": "7585df2fd9672883a1500e76331661f838d543263097041bee9efaf33517c187",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 735000,
+  "step": 740000,
   "stats": {
-    "climbmix/blocks": 26460000,
-    "climbmix/targets": 44233581693,
-    "climbmix/input_tokens": 44236384555,
-    "ultra_qa/blocks": 9922500,
-    "ultra_qa/targets": 16187472584,
-    "ultra_qa/input_tokens": 16187503812,
-    "cortex/blocks": 9922500,
-    "cortex/targets": 377265616,
-    "cortex/input_tokens": 15772193657,
-    "ultra_style/blocks": 9922500,
-    "ultra_style/targets": 15694417562,
-    "ultra_style/input_tokens": 15695240794,
-    "openmath/blocks": 6615000,
-    "openmath/targets": 10221476545,
-    "openmath/input_tokens": 10224882264,
-    "rewrite6/blocks": 3307500,
-    "rewrite6/targets": 5422206078,
-    "rewrite6/input_tokens": 5428150619
+    "climbmix/blocks": 26640000,
+    "climbmix/targets": 44534380574,
+    "climbmix/input_tokens": 44537202395,
+    "ultra_qa/blocks": 9990000,
+    "ultra_qa/targets": 16297612094,
+    "ultra_qa/input_tokens": 16297643521,
+    "cortex/blocks": 9990000,
+    "cortex/targets": 379836958,
+    "cortex/input_tokens": 15879487851,
+    "ultra_style/blocks": 9990000,
+    "ultra_style/targets": 15801366047,
+    "ultra_style/input_tokens": 15802194940,
+    "openmath/blocks": 6660000,
+    "openmath/targets": 10291092474,
+    "openmath/input_tokens": 10294521411,
+    "rewrite6/blocks": 3330000,
+    "rewrite6/targets": 5459078580,
+    "rewrite6/input_tokens": 5465063978
   },
   "sources": {
     "climbmix": {
-      "page": 3623,
-      "row": 2520,
+      "page": 3648,
+      "row": 803,
       "start": {
-        "file": 175,
+        "file": 176,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 31,
+          "row_group": 48,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 14842328
+      "documents": 14943011
     },
     "rewrite6": {
-      "page": 11,
-      "row": 2259,
+      "page": 25,
+      "row": 3061,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 45056
+          "row": 102400
         }
       },
       "epoch": 41,
-      "offset": 0,
-      "documents": 8468879
+      "offset": 2048,
+      "documents": 8527025
     },
     "openmath": {
-      "page": 2264,
-      "row": 3335,
+      "page": 2280,
+      "row": 890,
       "start": {
         "file": 21,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 103,
-          "row": 694
+          "row_group": 169,
+          "row": 230
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 9276679
+      "documents": 9339770
     },
     "ultra_qa": {
-      "page": 979,
-      "row": 1035,
+      "page": 985,
+      "row": 3725,
       "start": {
         "file": 7,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 49879
+          "row_group": 6,
+          "row": 9755
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 4011019
+      "offset": 0,
+      "documents": 4038285
     },
     "ultra_style": {
-      "page": 1494,
-      "row": 2372,
+      "page": 1504,
+      "row": 3284,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 43777
+          "row": 84737
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 6121796
+      "documents": 6163668
     },
     "cortex": {
-      "page": 192433,
-      "row": 128,
+      "page": 193744,
+      "row": 26,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 24631552
+      "documents": 24799258
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 735000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315", "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "541a1f3f705d2b085cbb51195f8b00accd55d018941e78dc4445045fc4c03dcd", "coverage.json": "fe7d1dc256a120fa7e42273b21198470f2acfeaaf06e57edd8a92aeb71326f7c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000735000"}
+{"step": 740000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393", "training_config.json": "a5b628e58d936dba526336e6ed1a3b16a703b7373430cc4aca66220c358ed037", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "7f8712a7f9538f373e624550e1f5a9ce3a8e415d6a140e719f8256eddac3ad48", "coverage.json": "7585df2fd9672883a1500e76331661f838d543263097041bee9efaf33517c187", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000740000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1c6719e35478e41b41395590fac06c68dd51c2c2fe17fc071f3f0d62497ad73
+oid sha256:965f2a1b8062278f91a7a579c0f9910dd996acdaaa15c0589449d5be638afbb9
 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": 76078.75752915804,
-      "required_pages_per_second_upper_estimate": 0.0052633049729053005
+      "required_token_positions_per_second": 75740.10139994962,
+      "required_pages_per_second_upper_estimate": 0.005239875956096181
     },
     {
       "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": 9509.844691144755,
-      "required_pages_per_second_upper_estimate": 0.003981573384042948
+      "required_token_positions_per_second": 9467.512674993703,
+      "required_pages_per_second_upper_estimate": 0.003963849852873521
     },
     {
       "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": 19019.68938228951,
-      "required_pages_per_second_upper_estimate": 0.0039612997153919026
+      "required_token_positions_per_second": 18935.025349987405,
+      "required_pages_per_second_upper_estimate": 0.0039436664302040505
     },
     {
       "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": 28529.534073434268,
-      "required_pages_per_second_upper_estimate": 0.0017103020056261802
+      "required_token_positions_per_second": 28402.538024981106,
+      "required_pages_per_second_upper_estimate": 0.0017026887864331514
     },
     {
       "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": 28529.534073434268,
-      "required_pages_per_second_upper_estimate": 0.0022180457860511227
+      "required_token_positions_per_second": 28402.538024981106,
+      "required_pages_per_second_upper_estimate": 0.002208172401880472
     },
     {
       "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": 28529.534073434268,
-      "required_pages_per_second_upper_estimate": 0.36594623046695485
+      "required_token_positions_per_second": 28402.538024981106,
+      "required_pages_per_second_upper_estimate": 0.3643172615151307
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.08382008823436138,
-    "producer_p95_seconds": 0.23780670700034534,
-    "gpu_update_seconds": 0.9691009999965221,
-    "producer_batches_per_second": 11.930314332335168,
-    "gpu_batches_per_second": 1.031884189577339,
-    "producer_headroom": 11.561679549738852,
+    "producer_batch_seconds": 0.11142892917197855,
+    "producer_p95_seconds": 0.2649414719999186,
+    "gpu_update_seconds": 0.973434133797569,
+    "producer_batches_per_second": 8.974330162112638,
+    "gpu_batches_per_second": 1.0272908718526153,
+    "producer_headroom": 8.735919307769514,
     "sustainable": true,
-    "buffer_seconds": 60.084261999784374,
+    "buffer_seconds": 60.35291629544928,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 304411.78655841085,
-    "upload_bytes_per_second": 4856659.550780564,
+    "generation_bytes_per_second": 287306.2932266875,
+    "upload_bytes_per_second": 5098154.501941984,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 65.319683051722,
+    "estimated_full_shards_per_hour": 61.64924237823934,
     "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": 1.54305899999963,
-      "bytes": 5745307,
-      "pages": 28,
+      "seconds": 1.2874604959997669,
+      "bytes": 2530940,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 0.9886994910002613,
-      "bytes": 1117526,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.8489439929999207,
-      "bytes": 17752294,
-      "pages": 35,
-      "shards": 5
+      "seconds": 1.3693337599997903,
+      "bytes": 10798536,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.054843687000357,
-      "bytes": 1282245,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.092897812000956,
+      "bytes": 6537422,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.0156906100000924,
-      "bytes": 5808038,
-      "pages": 28,
+      "seconds": 0.9582660920004855,
+      "bytes": 2603502,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.466896263000308,
-      "bytes": 7685001,
-      "pages": 38,
+      "seconds": 1.5727497110001423,
+      "bytes": 11757823,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.0174859720000313,
-      "bytes": 4930803,
-      "pages": 30,
+      "seconds": 1.1941701559990179,
+      "bytes": 5221736,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.3428281450014765,
-      "bytes": 7443989,
-      "pages": 38,
+      "seconds": 1.6056268569991516,
+      "bytes": 7578967,
+      "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.0973035269998945,
-      "bytes": 6756483,
-      "pages": 32,
+      "seconds": 1.20693444599965,
+      "bytes": 6453330,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.110015972999463,
-      "bytes": 6170503,
-      "pages": 36,
+      "seconds": 0.9651811909989192,
+      "bytes": 2682508,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.4518775689994072,
-      "bytes": 12256015,
-      "pages": 37,
-      "shards": 4
+      "seconds": 1.4213833419998991,
+      "bytes": 10754179,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 0.9289161779997812,
-      "bytes": 848234,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.5129684479998105,
+      "bytes": 6722253,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.4199929900005372,
-      "bytes": 7653735,
+      "seconds": 1.165229547999843,
+      "bytes": 7984438,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 3.9366146709999157,
-      "bytes": 6393409,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.001051560999258,
+      "bytes": 939443,
+      "pages": 30,
+      "shards": 1
     },
     {
-      "seconds": 1.550046686999849,
-      "bytes": 10810645,
-      "pages": 27,
-      "shards": 3
+      "seconds": 1.4753997340012575,
+      "bytes": 11016252,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.2842068059999292,
-      "bytes": 7447718,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.6974678700007644,
+      "bytes": 6266676,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3457034299990482,
-      "bytes": 1181953,
-      "pages": 35,
+      "seconds": 1.6492898739998054,
+      "bytes": 1094803,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.210031186001288,
-      "bytes": 5583238,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.8326161909990333,
+      "bytes": 13393499,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.2692713039996306,
-      "bytes": 11755459,
+      "seconds": 1.4041516550005326,
+      "bytes": 11034703,
       "pages": 37,
-      "shards": 4
+      "shards": 3
     },
     {
-      "seconds": 1.0984560849992704,
-      "bytes": 1168159,
-      "pages": 35,
+      "seconds": 1.1081106500005262,
+      "bytes": 1056276,
+      "pages": 28,
       "shards": 1
     },
     {
-      "seconds": 1.9164282879992243,
-      "bytes": 13374744,
-      "pages": 37,
-      "shards": 5
+      "seconds": 1.211192590999417,
+      "bytes": 7370118,
+      "pages": 35,
+      "shards": 3
     },
     {
-      "seconds": 0.928293617998861,
-      "bytes": 992050,
-      "pages": 33,
+      "seconds": 1.246054737999657,
+      "bytes": 1155172,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.3297191130004649,
-      "bytes": 10315993,
+      "seconds": 1.6413428979994933,
+      "bytes": 11205821,
       "pages": 33,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.4511640289983916,
-      "bytes": 7595757,
+      "seconds": 1.3393069079993438,
+      "bytes": 11931034,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.2930020470012096,
-      "bytes": 547633,
-      "pages": 17,
+      "seconds": 0.9887657250001212,
+      "bytes": 1035853,
+      "pages": 34,
       "shards": 1
     },
     {
-      "seconds": 1.3445512510006665,
-      "bytes": 5828365,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.3170902919991931,
+      "bytes": 7309091,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 1.2113412039998366,
-      "bytes": 6839007,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.3663103399994725,
+      "bytes": 5168575,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.4825698250006099,
-      "bytes": 11644292,
-      "pages": 29,
-      "shards": 3
+      "seconds": 1.1543622389999655,
+      "bytes": 6027007,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 1.3249330049984565,
-      "bytes": 7496918,
+      "seconds": 1.39643492399955,
+      "bytes": 7665740,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.1885178210013692,
-      "bytes": 1081987,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.1344610849992023,
+      "bytes": 6680190,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.9165967070002807,
-      "bytes": 10316454,
+      "seconds": 1.5331677430003765,
+      "bytes": 11639669,
       "pages": 33,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.1245960120013478,
-      "bytes": 2666345,
-      "pages": 37,
-      "shards": 2
+      "seconds": 0.9482507379998424,
+      "bytes": 1158257,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.2064912770001683,
-      "bytes": 6004944,
-      "pages": 31,
+      "seconds": 1.2962373409991415,
+      "bytes": 6282079,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 3.7895365029999084,
-      "bytes": 13350183,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.4015076579999004,
+      "bytes": 7431338,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.1515760519996547,
-      "bytes": 5491935,
-      "pages": 30,
+      "seconds": 0.9902176969990251,
+      "bytes": 5564439,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1848105680001026,
-      "bytes": 6290344,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.667742324003484,
+      "bytes": 13340513,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.0463193560008222,
-      "bytes": 2621559,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2163241669986746,
+      "bytes": 1010699,
+      "pages": 28,
+      "shards": 1
     },
     {
-      "seconds": 1.171350075999726,
-      "bytes": 6139992,
-      "pages": 36,
+      "seconds": 1.3674872539995704,
+      "bytes": 6041473,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.280977876000179,
-      "bytes": 9975922,
+      "seconds": 1.3613106869997864,
+      "bytes": 11006649,
       "pages": 33,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.2528174399994896,
-      "bytes": 8102539,
-      "pages": 30,
-      "shards": 3
+      "seconds": 0.8713928039978782,
+      "bytes": 1169048,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.1965843840007437,
-      "bytes": 6380308,
-      "pages": 36,
+      "seconds": 1.0586420059989905,
+      "bytes": 6189631,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2874604959997669,
-      "bytes": 2530940,
-      "pages": 33,
-      "shards": 2
+      "seconds": 1.3973069029998442,
+      "bytes": 13145960,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 1.3693337599997903,
-      "bytes": 10798536,
-      "pages": 37,
+      "seconds": 1.3652075969985162,
+      "bytes": 5700613,
+      "pages": 36,
       "shards": 3
     },
     {
-      "seconds": 1.092897812000956,
-      "bytes": 6537422,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.7661054199998034,
+      "bytes": 7454533,
+      "pages": 30,
+      "shards": 3
     },
     {
-      "seconds": 0.9582660920004855,
-      "bytes": 2603502,
+      "seconds": 0.9842067560020951,
+      "bytes": 1088907,
+      "pages": 33,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1375966550003795,
+      "bytes": 6241738,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.5727497110001423,
-      "bytes": 11757823,
-      "pages": 31,
-      "shards": 3
+      "seconds": 1.329601728997659,
+      "bytes": 12006565,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1941701559990179,
-      "bytes": 5221736,
-      "pages": 34,
+      "seconds": 1.637713142998109,
+      "bytes": 6812710,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.6056268569991516,
-      "bytes": 7578967,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.20693444599965,
-      "bytes": 6453330,
-      "pages": 35,
+      "seconds": 1.216705439001089,
+      "bytes": 6221386,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9651811909989192,
-      "bytes": 2682508,
-      "pages": 37,
+      "seconds": 1.2352482690002944,
+      "bytes": 2570616,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.4213833419998991,
-      "bytes": 10754179,
-      "pages": 34,
+      "seconds": 1.360712293000688,
+      "bytes": 10168781,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 1.5129684479998105,
-      "bytes": 6722253,
+      "seconds": 1.3043567909990088,
+      "bytes": 7359704,
       "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.165229547999843,
-      "bytes": 7984438,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.5744073349997052,
+      "bytes": 1306429,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.001051560999258,
-      "bytes": 939443,
-      "pages": 30,
-      "shards": 1
+      "seconds": 1.5331175120008993,
+      "bytes": 11738494,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.4753997340012575,
-      "bytes": 11016252,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0237298360007117,
+      "bytes": 7075351,
+      "pages": 32,
+      "shards": 3
     },
     {
-      "seconds": 1.6974678700007644,
-      "bytes": 6266676,
-      "pages": 36,
+      "seconds": 1.4313278709996666,
+      "bytes": 6586683,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.6492898739998054,
-      "bytes": 1094803,
-      "pages": 34,
+      "seconds": 1.1789775520010153,
+      "bytes": 937197,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.8326161909990333,
-      "bytes": 13393499,
-      "pages": 34,
-      "shards": 4
+      "seconds": 1.0937288770001032,
+      "bytes": 7693834,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.4041516550005326,
-      "bytes": 11034703,
-      "pages": 37,
+      "seconds": 1.3860260590008693,
+      "bytes": 10598757,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.1081106500005262,
-      "bytes": 1056276,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.0763569690025179,
+      "bytes": 2341572,
+      "pages": 29,
+      "shards": 2
     },
     {
-      "seconds": 1.211192590999417,
-      "bytes": 7370118,
-      "pages": 35,
+      "seconds": 1.3760699319973355,
+      "bytes": 11906705,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.246054737999657,
-      "bytes": 1155172,
+      "seconds": 1.2023094430005585,
+      "bytes": 5878390,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9879657829987991,
+      "bytes": 6822345,
       "pages": 34,
-      "shards": 1
+      "shards": 3
     },
     {
-      "seconds": 1.6413428979994933,
-      "bytes": 11205821,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.0887722430015856,
+      "bytes": 6046508,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.3393069079993438,
-      "bytes": 11931034,
-      "pages": 38,
-      "shards": 3
+      "seconds": 0.9664684920026048,
+      "bytes": 1011694,
+      "pages": 29,
+      "shards": 1
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ba531a5375e7fdc02ffb63552e7ec706eb03a955e5cb37672bc7eb52d7cc315
-size 721775167
+oid sha256:7a080253df8bf544a1d7795cc94dbd9bed18537ff0c92c322418f47ac23af393
+size 721766975