Qwen90M step 865000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 860000,
+  "step": 865000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0",
+    "model.safetensors": "f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd",
-    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
+    "trainer.pt": "18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9",
+    "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58",
     "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367",
-    "shard_scheduler.json": "9a29ea4bdc9cefd2e92ebd24e57d2572330e585a6e74a99d3c512c6e2de2f678",
-    "coverage.json": "7fba50c91c135d7273f46b84ac1af03e10b40c0850608fe8d645d5160ea5fc41",
+    "shard_scheduler.json": "c95b4d546c5df7ee17ed03b0bd9cd6934d744393d9e9e9817ef0f64a430ddf94",
+    "coverage.json": "3109f65091eb186c8eaa80402ee04bc03abb10fd556167b2666aaef70ee57e60",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 860000,
+  "step": 865000,
   "stats": {
-    "climbmix/blocks": 30960000,
-    "climbmix/targets": 51754676559,
-    "climbmix/input_tokens": 51757956002,
-    "ultra_qa/blocks": 11610000,
-    "ultra_qa/targets": 18941056864,
-    "ultra_qa/input_tokens": 18941093290,
-    "cortex/blocks": 11610000,
-    "cortex/targets": 441483304,
-    "cortex/input_tokens": 18454886279,
-    "ultra_style/blocks": 11610000,
-    "ultra_style/targets": 18363767524,
-    "ultra_style/input_tokens": 18364730231,
-    "openmath/blocks": 7740000,
-    "openmath/targets": 11959461269,
-    "openmath/input_tokens": 11963446369,
-    "rewrite6/blocks": 3870000,
-    "rewrite6/targets": 6344361516,
-    "rewrite6/input_tokens": 6351317040
+    "climbmix/blocks": 31140000,
+    "climbmix/targets": 52055877655,
+    "climbmix/input_tokens": 52059176107,
+    "ultra_qa/blocks": 11677500,
+    "ultra_qa/targets": 19051109408,
+    "ultra_qa/input_tokens": 19051146044,
+    "cortex/blocks": 11677500,
+    "cortex/targets": 444043754,
+    "cortex/input_tokens": 18562155884,
+    "ultra_style/blocks": 11677500,
+    "ultra_style/targets": 18470831928,
+    "ultra_style/input_tokens": 18471800000,
+    "openmath/blocks": 7785000,
+    "openmath/targets": 12029146241,
+    "openmath/input_tokens": 12033154511,
+    "rewrite6/blocks": 3892500,
+    "rewrite6/targets": 6381217228,
+    "rewrite6/input_tokens": 6388213504
   },
   "sources": {
     "climbmix": {
-      "page": 4240,
-      "row": 396,
+      "page": 4264,
+      "row": 2484,
       "start": {
-        "file": 205,
+        "file": 206,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 20,
+          "row_group": 33,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 17367436
+      "offset": 2048,
+      "documents": 17467828
     },
     "rewrite6": {
-      "page": 12,
-      "row": 642,
+      "page": 26,
+      "row": 1330,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 49152
+          "row": 106496
         }
       },
       "epoch": 48,
       "offset": 2048,
-      "documents": 9909186
+      "documents": 9967218
     },
     "openmath": {
-      "page": 2650,
-      "row": 161,
+      "page": 2665,
+      "row": 1871,
       "start": {
         "file": 24,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 374,
-          "row": 801
+          "row_group": 436,
+          "row": 241
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 10854561
+      "offset": 2048,
+      "documents": 10917711
     },
     "ultra_qa": {
-      "page": 1145,
-      "row": 3108,
+      "page": 1152,
+      "row": 1712,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 13278
+          "row": 41950
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 4693028
+      "offset": 0,
+      "documents": 4720304
     },
     "ultra_style": {
-      "page": 1748,
-      "row": 2510,
+      "page": 1758,
+      "row": 2694,
       "start": {
         "file": 10,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 3,
-          "row": 55900
+          "row_group": 4,
+          "row": 12042
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 7162318
+      "offset": 0,
+      "documents": 7203462
     },
     "cortex": {
-      "page": 225159,
-      "row": 87,
+      "page": 226468,
+      "row": 78,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 28820439
+      "documents": 28987982
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 860000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "9a29ea4bdc9cefd2e92ebd24e57d2572330e585a6e74a99d3c512c6e2de2f678", "coverage.json": "7fba50c91c135d7273f46b84ac1af03e10b40c0850608fe8d645d5160ea5fc41", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000860000"}
+{"step": 865000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9", "training_config.json": "0ab64376a29947d9a9f670c0ece06c1900ab6f9ec82482957af5fc2c70beec58", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "c95b4d546c5df7ee17ed03b0bd9cd6934d744393d9e9e9817ef0f64a430ddf94", "coverage.json": "3109f65091eb186c8eaa80402ee04bc03abb10fd556167b2666aaef70ee57e60", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000865000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0
+oid sha256:f24ee40d83e4992cf8ce8af69aca975058968fa55d27163a7767a086b92a8387
 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": 76235.28940132579,
-      "required_pages_per_second_upper_estimate": 0.0052741342110258485
+      "required_token_positions_per_second": 75711.298881415,
+      "required_pages_per_second_upper_estimate": 0.0052378833310329055
     },
     {
       "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": 9529.411175165724,
-      "required_pages_per_second_upper_estimate": 0.003989765462307878
+      "required_token_positions_per_second": 9463.912360176875,
+      "required_pages_per_second_upper_estimate": 0.003962342476242838
     },
     {
       "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": 19058.822350331448,
-      "required_pages_per_second_upper_estimate": 0.003969450080629271
+      "required_token_positions_per_second": 18927.82472035375,
+      "required_pages_per_second_upper_estimate": 0.003942166728944734
     },
     {
       "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": 28588.23352549717,
-      "required_pages_per_second_upper_estimate": 0.0017138209481484775
+      "required_token_positions_per_second": 28391.737080530624,
+      "required_pages_per_second_upper_estimate": 0.0017020412863054321
     },
     {
       "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": 28588.23352549717,
-      "required_pages_per_second_upper_estimate": 0.00222260941025741
+      "required_token_positions_per_second": 28391.737080530624,
+      "required_pages_per_second_upper_estimate": 0.002207332675957781
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.04061180899998362,
       "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": 28588.23352549717,
-      "required_pages_per_second_upper_estimate": 0.36669916401145664
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28391.737080530624,
+      "required_pages_per_second_upper_estimate": 0.3641787185968705
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,413 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 63,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.06858718393758068,
-    "producer_p95_seconds": 0.22669146000043838,
-    "gpu_update_seconds": 0.9671111709417584,
-    "producer_batches_per_second": 14.579983352430284,
-    "gpu_batches_per_second": 1.0340072889719751,
-    "producer_headroom": 14.100464772280196,
+    "producer_batch_seconds": 0.13520461964066044,
+    "producer_p95_seconds": 0.2855433209988405,
+    "gpu_update_seconds": 0.9738044530906622,
+    "producer_batches_per_second": 7.396196983932548,
+    "gpu_batches_per_second": 1.0269002126928033,
+    "producer_headroom": 7.202449558889239,
     "sustainable": true,
-    "buffer_seconds": 60.92800376933078,
+    "buffer_seconds": 60.37587609162106,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 77,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 389341.2469068264,
-    "upload_bytes_per_second": 3447999.215230209,
+    "generation_bytes_per_second": 358001.55127571075,
+    "upload_bytes_per_second": 3629836.3380909152,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 83.54356818584054,
+    "estimated_full_shards_per_hour": 76.81879905417911,
     "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": 1.182235641999796,
-      "bytes": 6603515,
-      "pages": 27,
+      "seconds": 1.716705255999841,
+      "bytes": 6235687,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.285518773000149,
-      "bytes": 6002809,
-      "pages": 36,
-      "shards": 2
+      "seconds": 33.88602663399979,
+      "bytes": 7355362,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.7619018259993027,
-      "bytes": 7674402,
-      "pages": 38,
+      "seconds": 1.5573600560001069,
+      "bytes": 11648033,
+      "pages": 46,
       "shards": 3
     },
     {
-      "seconds": 1.6144984039983683,
-      "bytes": 4728422,
-      "pages": 30,
-      "shards": 2
+      "seconds": 1.2928704839996499,
+      "bytes": 7315946,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.6315553869990254,
-      "bytes": 5703905,
-      "pages": 24,
+      "seconds": 1.4081739580014982,
+      "bytes": 5891309,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.11825536699871,
-      "bytes": 2656311,
-      "pages": 35,
+      "seconds": 1.286470338000072,
+      "bytes": 2778002,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.9071076060008636,
-      "bytes": 12136915,
+      "seconds": 1.1679330120005034,
+      "bytes": 10820666,
       "pages": 37,
       "shards": 3
     },
     {
-      "seconds": 1.7720820229988021,
-      "bytes": 11800483,
-      "pages": 32,
-      "shards": 4
-    },
-    {
-      "seconds": 1.0689491579996684,
-      "bytes": 1154476,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.4827527400011604,
+      "bytes": 6647922,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.7027444539999124,
-      "bytes": 6185388,
-      "pages": 36,
-      "shards": 2
+      "seconds": 1.248109399999521,
+      "bytes": 7422482,
+      "pages": 37,
+      "shards": 3
     },
     {
-      "seconds": 1.5859740610012523,
-      "bytes": 2563401,
-      "pages": 35,
+      "seconds": 1.451942487999986,
+      "bytes": 6077412,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.2778493570003775,
-      "bytes": 10237665,
-      "pages": 33,
+      "seconds": 1.4631003799986502,
+      "bytes": 7075109,
+      "pages": 43,
       "shards": 3
     },
     {
-      "seconds": 2.496101438000551,
-      "bytes": 13297971,
-      "pages": 33,
-      "shards": 4
+      "seconds": 2.129564337999909,
+      "bytes": 5884538,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 1.2942262079995999,
-      "bytes": 1101377,
-      "pages": 34,
-      "shards": 1
+      "seconds": 1.1728028199995606,
+      "bytes": 6147166,
+      "pages": 33,
+      "shards": 2
     },
     {
-      "seconds": 1.815276637998977,
-      "bytes": 10307802,
-      "pages": 39,
+      "seconds": 1.2317090710002958,
+      "bytes": 8202394,
+      "pages": 32,
       "shards": 3
     },
     {
-      "seconds": 1.1417645319997973,
-      "bytes": 2370286,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.6838970150001842,
+      "bytes": 9992623,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 1.1171746460004215,
-      "bytes": 6111968,
+      "seconds": 1.2029927740004496,
+      "bytes": 6068424,
       "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.2683859379994828,
-      "bytes": 7618531,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.1870869580016006,
+      "bytes": 2666964,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.7501807870012271,
-      "bytes": 11262011,
-      "pages": 28,
+      "seconds": 1.4462591800001974,
+      "bytes": 10549433,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.2981046400000196,
-      "bytes": 6199716,
+      "seconds": 1.0944702749984572,
+      "bytes": 2560030,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.6966266469989932,
-      "bytes": 7696834,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.3531785280028998,
+      "bytes": 11787584,
+      "pages": 32,
+      "shards": 4
     },
     {
-      "seconds": 1.3060271180002019,
-      "bytes": 1241981,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.0967427190007584,
+      "bytes": 6096101,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.5251753980010108,
-      "bytes": 9924170,
-      "pages": 32,
+      "seconds": 1.3933399300003657,
+      "bytes": 2445888,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.9226194889997714,
+      "bytes": 9744014,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.1737768090006284,
-      "bytes": 7614180,
+      "seconds": 1.1809931960015092,
+      "bytes": 7571529,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.2979090160006308,
-      "bytes": 6818365,
-      "pages": 30,
-      "shards": 2
+      "seconds": 0.7855844450023142,
+      "bytes": 302035,
+      "pages": 9,
+      "shards": 1
     },
     {
-      "seconds": 0.9478092970002763,
-      "bytes": 253929,
-      "pages": 8,
+      "seconds": 0.93341918599981,
+      "bytes": 1061744,
+      "pages": 32,
       "shards": 1
     },
     {
-      "seconds": 1.6955216110000038,
-      "bytes": 7314599,
-      "pages": 30,
-      "shards": 3
+      "seconds": 1.5622991370000818,
+      "bytes": 16646339,
+      "pages": 40,
+      "shards": 4
     },
     {
-      "seconds": 1.2749757900000986,
-      "bytes": 5237846,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.5330627459989046,
+      "bytes": 7305115,
+      "pages": 34,
+      "shards": 3
     },
     {
-      "seconds": 1.5798416210000141,
-      "bytes": 6237294,
-      "pages": 36,
-      "shards": 2
+      "seconds": 0.8943780709996645,
+      "bytes": 971661,
+      "pages": 29,
+      "shards": 1
     },
     {
-      "seconds": 1.5424033700001019,
-      "bytes": 7705972,
-      "pages": 39,
+      "seconds": 1.4041120760011836,
+      "bytes": 7469317,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.9795733470000414,
-      "bytes": 923741,
-      "pages": 28,
-      "shards": 2
+      "seconds": 1.2255340310002794,
+      "bytes": 10181330,
+      "pages": 31,
+      "shards": 3
     },
     {
-      "seconds": 1.5211821389984834,
-      "bytes": 18268501,
-      "pages": 39,
-      "shards": 5
+      "seconds": 1.0713620990027266,
+      "bytes": 1066838,
+      "pages": 34,
+      "shards": 1
     },
     {
-      "seconds": 1.3536671069996373,
-      "bytes": 6151717,
+      "seconds": 1.2431204950007668,
+      "bytes": 13191953,
       "pages": 36,
-      "shards": 2
+      "shards": 4
     },
     {
-      "seconds": 1.1601749330002349,
-      "bytes": 883373,
-      "pages": 25,
+      "seconds": 1.2758062670000072,
+      "bytes": 1083353,
+      "pages": 30,
       "shards": 1
     },
     {
-      "seconds": 1.1801181450009608,
-      "bytes": 7664644,
+      "seconds": 1.7504430440021679,
+      "bytes": 10417035,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.0472080479994474,
-      "bytes": 5473896,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.548700262001148,
-      "bytes": 13538856,
-      "pages": 41,
-      "shards": 4
+      "seconds": 1.7296301900023536,
+      "bytes": 7708812,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 1.3191421729989088,
-      "bytes": 5997715,
-      "pages": 32,
-      "shards": 2
+      "seconds": 1.0786898769983964,
+      "bytes": 798109,
+      "pages": 26,
+      "shards": 1
     },
     {
-      "seconds": 1.765130539000893,
-      "bytes": 937778,
+      "seconds": 1.6809108209999977,
+      "bytes": 7259646,
       "pages": 29,
-      "shards": 1
+      "shards": 3
     },
     {
-      "seconds": 2.7900099470007262,
-      "bytes": 11587181,
-      "pages": 37,
+      "seconds": 1.6123849089999567,
+      "bytes": 16470150,
+      "pages": 45,
       "shards": 4
     },
     {
-      "seconds": 0.987754063000466,
-      "bytes": 1220161,
+      "seconds": 1.350964933000796,
+      "bytes": 1228633,
       "pages": 35,
       "shards": 1
     },
     {
-      "seconds": 1.716705255999841,
-      "bytes": 6235687,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 33.88602663399979,
-      "bytes": 7355362,
-      "pages": 31,
+      "seconds": 1.590518836001138,
+      "bytes": 7289840,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.5573600560001069,
-      "bytes": 11648033,
-      "pages": 46,
-      "shards": 3
+      "seconds": 0.990031227000145,
+      "bytes": 1215408,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 1.2928704839996499,
-      "bytes": 7315946,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.4002689310000278,
+      "bytes": 11880783,
+      "pages": 32,
+      "shards": 4
     },
     {
-      "seconds": 1.4081739580014982,
-      "bytes": 5891309,
-      "pages": 28,
+      "seconds": 1.1156020849994093,
+      "bytes": 6256874,
+      "pages": 38,
       "shards": 2
     },
     {
-      "seconds": 1.286470338000072,
-      "bytes": 2778002,
-      "pages": 37,
+      "seconds": 1.0478186670006835,
+      "bytes": 6856744,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.1679330120005034,
-      "bytes": 10820666,
-      "pages": 37,
+      "seconds": 1.187728501001402,
+      "bytes": 7914573,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.4827527400011604,
-      "bytes": 6647922,
-      "pages": 30,
+      "seconds": 1.098511371998029,
+      "bytes": 5354193,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.248109399999521,
-      "bytes": 7422482,
-      "pages": 37,
+      "seconds": 2.7016539569995075,
+      "bytes": 7555720,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.451942487999986,
-      "bytes": 6077412,
-      "pages": 30,
+      "seconds": 1.483174205000978,
+      "bytes": 5900678,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.4631003799986502,
-      "bytes": 7075109,
-      "pages": 43,
-      "shards": 3
+      "seconds": 0.8595360259969311,
+      "bytes": 1182793,
+      "pages": 35,
+      "shards": 1
     },
     {
-      "seconds": 2.129564337999909,
-      "bytes": 5884538,
-      "pages": 31,
-      "shards": 2
+      "seconds": 1.7856543759990018,
+      "bytes": 17299146,
+      "pages": 27,
+      "shards": 6
     },
     {
-      "seconds": 1.1728028199995606,
-      "bytes": 6147166,
-      "pages": 33,
+      "seconds": 0.9336543670033279,
+      "bytes": 1145946,
+      "pages": 36,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1468644119995588,
+      "bytes": 6284755,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.2317090710002958,
-      "bytes": 8202394,
-      "pages": 32,
-      "shards": 3
+      "seconds": 1.0925358679996862,
+      "bytes": 2671177,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.6838970150001842,
-      "bytes": 9992623,
-      "pages": 38,
+      "seconds": 1.42488349499763,
+      "bytes": 9837963,
+      "pages": 31,
       "shards": 3
     },
     {
-      "seconds": 1.2029927740004496,
-      "bytes": 6068424,
-      "pages": 37,
-      "shards": 2
+      "seconds": 1.2818151600004057,
+      "bytes": 7876217,
+      "pages": 40,
+      "shards": 3
     },
     {
-      "seconds": 1.1870869580016006,
-      "bytes": 2666964,
-      "pages": 36,
+      "seconds": 1.1551925370004028,
+      "bytes": 6771836,
+      "pages": 30,
       "shards": 2
     },
     {
-      "seconds": 1.4462591800001974,
-      "bytes": 10549433,
-      "pages": 31,
+      "seconds": 1.460754736999661,
+      "bytes": 10771768,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.0944702749984572,
-      "bytes": 2560030,
-      "pages": 36,
+      "seconds": 0.9620606520002184,
+      "bytes": 2617341,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.3531785280028998,
-      "bytes": 11787584,
-      "pages": 32,
-      "shards": 4
+      "seconds": 1.1643879789990024,
+      "bytes": 6176643,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.0967427190007584,
-      "bytes": 6096101,
-      "pages": 36,
+      "seconds": 1.7241964509994432,
+      "bytes": 5999571,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.3933399300003657,
-      "bytes": 2445888,
-      "pages": 28,
+      "seconds": 1.129534389001492,
+      "bytes": 2663790,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.9226194889997714,
-      "bytes": 9744014,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.8240933249981026,
+      "bytes": 16162613,
+      "pages": 37,
+      "shards": 4
+    },
+    {
+      "seconds": 1.3452121949994762,
+      "bytes": 2460437,
+      "pages": 29,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd
-size 721770367
+oid sha256:18178f6d7c97fbaf7be1e689601dba5e41d5acf35448a4af6e2587866720a8d9
+size 721784575

training_config.json

@@ -53,7 +53,7 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
+  "prefetch_batches": 62,
   "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,