Qwen90M step 550000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 545000,
+  "step": 550000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b",
+    "model.safetensors": "2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7",
+    "trainer.pt": "76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead",
     "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb",
     "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
-    "shard_scheduler.json": "9362a296b2832b4b5007ca3ec1a25ba07f94fdb1146f86932cdf2269d63a6f4c",
-    "coverage.json": "682a00d4779d7723fea633cd98c833ad848bbb3df21fa445fa4095cd4ab9203f",
+    "shard_scheduler.json": "0714cd433019a43dac02d272ee7a38fe3cc84979e30ed3a971db3c2b11c513a0",
+    "coverage.json": "e56a23246a1ce8c2d499577bf18fa4049c8262287fd5f510b2a1aa7259f1d255",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 545000,
+  "step": 550000,
   "stats": {
-    "climbmix/blocks": 19620000,
-    "climbmix/targets": 32797131226,
-    "climbmix/input_tokens": 32799209443,
-    "ultra_qa/blocks": 7357500,
-    "ultra_qa/targets": 12002847915,
-    "ultra_qa/input_tokens": 12002871095,
-    "cortex/blocks": 7357500,
-    "cortex/targets": 279804176,
-    "cortex/input_tokens": 11695457470,
-    "ultra_style/blocks": 7357500,
-    "ultra_style/targets": 11636722703,
-    "ultra_style/input_tokens": 11637332832,
-    "openmath/blocks": 4905000,
-    "openmath/targets": 7578975444,
-    "openmath/input_tokens": 7581499711,
-    "rewrite6/blocks": 2452500,
-    "rewrite6/targets": 4020473427,
-    "rewrite6/input_tokens": 4024881624
+    "climbmix/blocks": 19800000,
+    "climbmix/targets": 33097835451,
+    "climbmix/input_tokens": 33099932886,
+    "ultra_qa/blocks": 7425000,
+    "ultra_qa/targets": 12112854779,
+    "ultra_qa/input_tokens": 12112878132,
+    "cortex/blocks": 7425000,
+    "cortex/targets": 282367717,
+    "cortex/input_tokens": 11802740868,
+    "ultra_style/blocks": 7425000,
+    "ultra_style/targets": 11743946324,
+    "ultra_style/input_tokens": 11744561998,
+    "openmath/blocks": 4950000,
+    "openmath/targets": 7648567050,
+    "openmath/input_tokens": 7651114390,
+    "rewrite6/blocks": 2475000,
+    "rewrite6/targets": 4057422202,
+    "rewrite6/input_tokens": 4061870651
   },
   "sources": {
     "climbmix": {
-      "page": 2687,
-      "row": 1151,
+      "page": 2711,
+      "row": 4062,
       "start": {
-        "file": 130,
+        "file": 131,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 3,
+          "row_group": 16,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 11007103
+      "documents": 11108318
     },
     "rewrite6": {
-      "page": 28,
-      "row": 3287,
+      "page": 42,
+      "row": 3348,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 114688
+          "row": 172032
         }
       },
       "epoch": 30,
       "offset": 0,
-      "documents": 6280095
+      "documents": 6337500
     },
     "openmath": {
-      "page": 1679,
-      "row": 155,
+      "page": 1694,
+      "row": 1687,
       "start": {
         "file": 15,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 327,
-          "row": 434
+          "row_group": 388,
+          "row": 874
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 6877339
+      "offset": 2048,
+      "documents": 6940311
     },
     "ultra_qa": {
-      "page": 726,
-      "row": 497,
+      "page": 732,
+      "row": 3118,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 5,
-          "row": 53043
+          "row_group": 6,
+          "row": 13034
         }
       },
       "epoch": 0,
       "offset": 4096,
-      "documents": 2974193
+      "documents": 3001390
     },
     "ultra_style": {
-      "page": 1108,
-      "row": 980,
+      "page": 1118,
+      "row": 1245,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 2462
+          "row": 43422
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 4539348
+      "offset": 4096,
+      "documents": 4580573
     },
     "cortex": {
-      "page": 142692,
-      "row": 98,
+      "page": 144001,
+      "row": 84,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 18264674
+      "documents": 18432212
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 545000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7", "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "9362a296b2832b4b5007ca3ec1a25ba07f94fdb1146f86932cdf2269d63a6f4c", "coverage.json": "682a00d4779d7723fea633cd98c833ad848bbb3df21fa445fa4095cd4ab9203f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000545000"}
+{"step": 550000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead", "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "0714cd433019a43dac02d272ee7a38fe3cc84979e30ed3a971db3c2b11c513a0", "coverage.json": "e56a23246a1ce8c2d499577bf18fa4049c8262287fd5f510b2a1aa7259f1d255", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000550000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b
+oid sha256:2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7
 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": 77724.66568028349,
-      "required_pages_per_second_upper_estimate": 0.00537717271783323
+      "required_token_positions_per_second": 77488.5880998427,
+      "required_pages_per_second_upper_estimate": 0.005360840323042881
     },
     {
       "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": 9715.583210035436,
-      "required_pages_per_second_upper_estimate": 0.004067711805593652
+      "required_token_positions_per_second": 9686.073512480338,
+      "required_pages_per_second_upper_estimate": 0.004055356711459891
     },
     {
       "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": 19431.166420070873,
-      "required_pages_per_second_upper_estimate": 0.00404699953098255
+      "required_token_positions_per_second": 19372.147024960675,
+      "required_pages_per_second_upper_estimate": 0.004034707347427211
     },
     {
       "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": 29146.74963010631,
-      "required_pages_per_second_upper_estimate": 0.0017473031358150829
+      "required_token_positions_per_second": 29058.220537441015,
+      "required_pages_per_second_upper_estimate": 0.001741995951885896
     },
     {
       "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": 29146.74963010631,
-      "required_pages_per_second_upper_estimate": 0.002266031580738054
+      "required_token_positions_per_second": 29058.220537441015,
+      "required_pages_per_second_upper_estimate": 0.002259148833181653
     },
     {
       "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": 29146.74963010631,
-      "required_pages_per_second_upper_estimate": 0.37386320891351205
+      "required_token_positions_per_second": 29058.220537441015,
+      "required_pages_per_second_upper_estimate": 0.37272765276793546
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.10329342065628566,
-    "producer_p95_seconds": 0.2151632269997208,
-    "gpu_update_seconds": 0.9485791846731952,
-    "producer_batches_per_second": 9.681158718981271,
-    "gpu_batches_per_second": 1.0542082476166923,
-    "producer_headroom": 9.18334564434305,
+    "producer_batch_seconds": 0.32388174496861666,
+    "producer_p95_seconds": 0.31020041200099513,
+    "gpu_update_seconds": 0.9514691363972556,
+    "producer_batches_per_second": 3.0875466602691595,
+    "gpu_batches_per_second": 1.051006240503509,
+    "producer_headroom": 2.937705354432528,
     "sustainable": true,
-    "buffer_seconds": 60.709067819084495,
+    "buffer_seconds": 60.89402472942436,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 297407.2945755357,
-    "upload_bytes_per_second": 5461795.829792473,
+    "generation_bytes_per_second": 458187.6926747766,
+    "upload_bytes_per_second": 3886870.9094786476,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 63.81668212842515,
+    "estimated_full_shards_per_hour": 98.31641278440927,
     "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": 1.1747847249989718,
-      "bytes": 10716254,
-      "pages": 34,
+      "seconds": 0.9234551270019438,
+      "bytes": 5872826,
+      "pages": 30,
+      "shards": 2
+    },
+    {
+      "seconds": 1.268712075001531,
+      "bytes": 10759229,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.0711626960001013,
-      "bytes": 2631436,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2720383849991777,
+      "bytes": 7578314,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.137612402000741,
-      "bytes": 5722821,
-      "pages": 28,
+      "seconds": 1.001895314999274,
+      "bytes": 6152237,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.4303736129986646,
-      "bytes": 11919616,
-      "pages": 42,
+      "seconds": 0.8757686259996262,
+      "bytes": 2112854,
+      "pages": 23,
       "shards": 3
     },
     {
-      "seconds": 1.0499967649993778,
-      "bytes": 7310677,
-      "pages": 35,
+      "seconds": 1.275124282001343,
+      "bytes": 10806597,
+      "pages": 46,
       "shards": 3
     },
     {
-      "seconds": 1.0670534790006059,
-      "bytes": 5802765,
-      "pages": 28,
-      "shards": 2
+      "seconds": 0.989914507001231,
+      "bytes": 672610,
+      "pages": 18,
+      "shards": 1
     },
     {
-      "seconds": 0.9111343770000531,
-      "bytes": 2649643,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.4889224110011128,
+      "bytes": 13780761,
+      "pages": 49,
+      "shards": 4
     },
     {
-      "seconds": 1.2196053769985156,
-      "bytes": 6057143,
-      "pages": 37,
+      "seconds": 0.993486828996538,
+      "bytes": 5717549,
+      "pages": 19,
       "shards": 2
     },
     {
-      "seconds": 1.2347321990000637,
-      "bytes": 9643472,
-      "pages": 35,
+      "seconds": 1.089157542999601,
+      "bytes": 7603521,
+      "pages": 46,
       "shards": 3
     },
     {
-      "seconds": 1.1648727040010272,
-      "bytes": 8469052,
+      "seconds": 1.0148978149991308,
+      "bytes": 6571585,
       "pages": 35,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1223673130007228,
-      "bytes": 6068911,
-      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1270621679996111,
-      "bytes": 5842224,
+      "seconds": 1.1422953030014469,
+      "bytes": 877918,
       "pages": 28,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.3927226420000807,
-      "bytes": 7163586,
-      "pages": 36,
-      "shards": 4
+      "seconds": 1.1798037480002677,
+      "bytes": 7431767,
+      "pages": 29,
+      "shards": 3
     },
     {
-      "seconds": 1.230830958000297,
-      "bytes": 6020552,
+      "seconds": 1.270509606001724,
+      "bytes": 15692090,
       "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2729082169989852,
-      "bytes": 7700586,
-      "pages": 39,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 0.9891643420014589,
-      "bytes": 6883017,
+      "seconds": 1.250901432998944,
+      "bytes": 1087271,
       "pages": 33,
-      "shards": 2
+      "shards": 1
     },
     {
-      "seconds": 1.2779713299987634,
-      "bytes": 9455856,
+      "seconds": 1.5110034520002955,
+      "bytes": 7289248,
       "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.159703276000073,
-      "bytes": 2570411,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2064628600001015,
-      "bytes": 6189523,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.0708622599995579,
+      "bytes": 1033241,
+      "pages": 32,
+      "shards": 1
     },
     {
-      "seconds": 1.4404870660000597,
-      "bytes": 11907122,
-      "pages": 40,
+      "seconds": 1.2405112279993773,
+      "bytes": 12607883,
+      "pages": 39,
       "shards": 4
     },
     {
-      "seconds": 0.842472636999446,
-      "bytes": 857193,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.1709360159984499,
+      "bytes": 6311701,
+      "pages": 35,
+      "shards": 2
     },
     {
-      "seconds": 1.0425618379995285,
-      "bytes": 11991809,
-      "pages": 35,
-      "shards": 3
+      "seconds": 1.1361314889982168,
+      "bytes": 6780795,
+      "pages": 32,
+      "shards": 2
     },
     {
-      "seconds": 1.543527111000003,
-      "bytes": 7770797,
-      "pages": 39,
+      "seconds": 1.0505489870010933,
+      "bytes": 7688939,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0592705330000172,
-      "bytes": 5143840,
+      "seconds": 1.4415530749975005,
+      "bytes": 9558241,
       "pages": 36,
-      "shards": 2
+      "shards": 3
     },
     {
-      "seconds": 0.9495372719993611,
-      "bytes": 5544847,
-      "pages": 18,
+      "seconds": 2.3017471879975346,
+      "bytes": 2579509,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.068828384999506,
-      "bytes": 2455135,
-      "pages": 29,
+      "seconds": 1.0516134489989781,
+      "bytes": 6165714,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.2464688429990929,
-      "bytes": 6118587,
-      "pages": 36,
+      "seconds": 0.908222386999114,
+      "bytes": 5648395,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.8108600380001008,
-      "bytes": 790179,
-      "pages": 24,
-      "shards": 1
-    },
-    {
-      "seconds": 2.05223027700049,
-      "bytes": 18642145,
-      "pages": 45,
-      "shards": 5
-    },
-    {
-      "seconds": 1.3427877739995893,
-      "bytes": 6044460,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.3117666090001876,
+      "bytes": 13264605,
+      "pages": 39,
+      "shards": 4
     },
     {
-      "seconds": 1.2501696180006547,
-      "bytes": 2425262,
+      "seconds": 1.0446096289997513,
+      "bytes": 5930135,
       "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.180065475000447,
-      "bytes": 6094205,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.9939750769990496,
+      "bytes": 1211971,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.2719860980032536,
-      "bytes": 10158142,
-      "pages": 41,
-      "shards": 3
+      "seconds": 1.65895906500009,
+      "bytes": 12493410,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 0.8902074819998234,
-      "bytes": 2207572,
-      "pages": 24,
-      "shards": 2
+      "seconds": 0.8486537320022762,
+      "bytes": 1121895,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.290630097999383,
-      "bytes": 12177821,
-      "pages": 42,
+      "seconds": 1.125895612000022,
+      "bytes": 7606699,
+      "pages": 33,
       "shards": 3
     },
     {
-      "seconds": 3.9100393570006418,
-      "bytes": 2234803,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9967152519966476,
-      "bytes": 11168227,
+      "seconds": 1.2285171910007193,
+      "bytes": 12124317,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0345216240020818,
-      "bytes": 6241760,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.8537721579996287,
+      "bytes": 1128490,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.262870265996753,
-      "bytes": 2380360,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.7903945020007086,
+      "bytes": 11075879,
+      "pages": 34,
+      "shards": 4
     },
     {
-      "seconds": 0.9234551270019438,
-      "bytes": 5872826,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.9962719150025805,
+      "bytes": 6075379,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.268712075001531,
-      "bytes": 10759229,
-      "pages": 40,
-      "shards": 3
+      "seconds": 0.8558462190012506,
+      "bytes": 876425,
+      "pages": 25,
+      "shards": 1
     },
     {
-      "seconds": 1.2720383849991777,
-      "bytes": 7578314,
+      "seconds": 1.5038929210022616,
+      "bytes": 12535029,
       "pages": 39,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.001895314999274,
-      "bytes": 6152237,
+      "seconds": 1.2274471419987094,
+      "bytes": 6988556,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.8757686259996262,
-      "bytes": 2112854,
-      "pages": 23,
+      "seconds": 1.103131982999912,
+      "bytes": 7144691,
+      "pages": 29,
       "shards": 3
     },
     {
-      "seconds": 1.275124282001343,
-      "bytes": 10806597,
-      "pages": 46,
-      "shards": 3
+      "seconds": 0.9400397869976587,
+      "bytes": 6119958,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 0.989914507001231,
-      "bytes": 672610,
-      "pages": 18,
+      "seconds": 0.879637284000637,
+      "bytes": 1034061,
+      "pages": 29,
       "shards": 1
     },
     {
-      "seconds": 1.4889224110011128,
-      "bytes": 13780761,
-      "pages": 49,
+      "seconds": 1.2181368159981503,
+      "bytes": 12003212,
+      "pages": 41,
       "shards": 4
     },
     {
-      "seconds": 0.993486828996538,
-      "bytes": 5717549,
-      "pages": 19,
-      "shards": 2
+      "seconds": 0.9120353730031638,
+      "bytes": 952497,
+      "pages": 31,
+      "shards": 1
     },
     {
-      "seconds": 1.089157542999601,
-      "bytes": 7603521,
-      "pages": 46,
-      "shards": 3
+      "seconds": 1.6672031760026584,
+      "bytes": 13248105,
+      "pages": 37,
+      "shards": 4
     },
     {
-      "seconds": 1.0148978149991308,
-      "bytes": 6571585,
-      "pages": 35,
+      "seconds": 1.4338108570009354,
+      "bytes": 5962058,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.1422953030014469,
-      "bytes": 877918,
-      "pages": 28,
-      "shards": 1
+      "seconds": 1.2595222179988923,
+      "bytes": 5231936,
+      "pages": 41,
+      "shards": 2
     },
     {
-      "seconds": 1.1798037480002677,
-      "bytes": 7431767,
-      "pages": 29,
+      "seconds": 1.146318651000911,
+      "bytes": 7213324,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.270509606001724,
-      "bytes": 15692090,
+      "seconds": 1.1155308010020235,
+      "bytes": 6292434,
       "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3688695919991005,
+      "bytes": 12940523,
+      "pages": 39,
       "shards": 4
     },
     {
-      "seconds": 1.250901432998944,
-      "bytes": 1087271,
-      "pages": 33,
+      "seconds": 1.308492041000136,
+      "bytes": 6108009,
+      "pages": 32,
+      "shards": 2
+    },
+    {
+      "seconds": 0.922122173000389,
+      "bytes": 1095644,
+      "pages": 31,
       "shards": 1
     },
     {
-      "seconds": 1.5110034520002955,
-      "bytes": 7289248,
-      "pages": 30,
+      "seconds": 1.0743960049985617,
+      "bytes": 7509044,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0708622599995579,
-      "bytes": 1033241,
-      "pages": 32,
+      "seconds": 1.3776889159998973,
+      "bytes": 10102958,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 4.997759090001637,
+      "bytes": 838514,
+      "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.2405112279993773,
-      "bytes": 12607883,
-      "pages": 39,
+      "seconds": 34.2643265609986,
+      "bytes": 13380891,
+      "pages": 44,
       "shards": 4
     },
     {
-      "seconds": 1.1709360159984499,
-      "bytes": 6311701,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.1768819699973392,
+      "bytes": 10254247,
+      "pages": 45,
+      "shards": 3
     },
     {
-      "seconds": 1.1361314889982168,
-      "bytes": 6780795,
-      "pages": 32,
+      "seconds": 2.0107650110003306,
+      "bytes": 2402302,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.0505489870010933,
-      "bytes": 7688939,
+      "seconds": 1.039769940001861,
+      "bytes": 6252309,
       "pages": 38,
-      "shards": 3
+      "shards": 2
     },
     {
-      "seconds": 1.4415530749975005,
-      "bytes": 9558241,
-      "pages": 36,
+      "seconds": 1.272125319999759,
+      "bytes": 7866146,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 2.3017471879975346,
-      "bytes": 2579509,
-      "pages": 37,
+      "seconds": 1.004612441000063,
+      "bytes": 5431394,
+      "pages": 26,
       "shards": 2
     },
     {
-      "seconds": 1.0516134489989781,
-      "bytes": 6165714,
-      "pages": 27,
-      "shards": 2
+      "seconds": 1.1393026930018095,
+      "bytes": 12031348,
+      "pages": 42,
+      "shards": 3
     },
     {
-      "seconds": 0.908222386999114,
-      "bytes": 5648395,
-      "pages": 35,
-      "shards": 2
+      "seconds": 1.1874672249978175,
+      "bytes": 7293811,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9514401040032681,
+      "bytes": 1189191,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3702037240000209,
+      "bytes": 10453891,
+      "pages": 36,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7
-size 721774207
+oid sha256:76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead
+size 721775551