Qwen90M step 940000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 935000,
+  "step": 940000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "32c5ec46c4d26de24f43e5e5ac475183bd0d5a9168b597e1a4995f06cda64d9b",
+    "model.safetensors": "d09a6f6d15b58b397b5cb80bf0442935d3486b863565dc4a3b414d4d29102f73",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "c55cec8e778846c453320337a4022807364361a6bb535fd24e51ca5d5f695ab5",
-    "training_config.json": "aafacc05aaf9ea874dcfa3d02457799efaf5525cf7b211a4208541f6e989efee",
+    "trainer.pt": "379eeb0d627e11d6aeb818be90d2beb7ad27d2bcf192919683768be57619bc1b",
+    "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4",
-    "shard_scheduler.json": "9525e5f4fb5139b10175901a17eb2c1ac2df509c25c3edcfbc3108aaabd5b46e",
-    "coverage.json": "ea1e9229096e365e0bf25587fc7c00524d8f83b2754ab83b1c551cc02c49d653",
+    "shard_scheduler.json": "e7f0f1e8787fc3b9f72fcf8f200650769ecf8ab43afd791855595c601ac93a27",
+    "coverage.json": "3d6b4fdad9262d21f1491946476e23ba4ec13ac73d7d5a5a0af7a9f9856d8a06",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 935000,
+  "step": 940000,
   "stats": {
-    "climbmix/blocks": 33660000,
-    "climbmix/targets": 56269025726,
-    "climbmix/input_tokens": 56272591695,
-    "ultra_qa/blocks": 12622500,
-    "ultra_qa/targets": 20592546150,
-    "ultra_qa/input_tokens": 20592585687,
-    "cortex/blocks": 12622500,
-    "cortex/targets": 479968870,
-    "cortex/input_tokens": 20064566894,
-    "ultra_style/blocks": 12622500,
-    "ultra_style/targets": 19964286118,
-    "ultra_style/input_tokens": 19965332428,
-    "openmath/blocks": 8415000,
-    "openmath/targets": 13002775878,
-    "openmath/input_tokens": 13007108390,
-    "rewrite6/blocks": 4207500,
-    "rewrite6/targets": 6897626953,
-    "rewrite6/input_tokens": 6905189624
+    "climbmix/blocks": 33840000,
+    "climbmix/targets": 56570146487,
+    "climbmix/input_tokens": 56573731401,
+    "ultra_qa/blocks": 12690000,
+    "ultra_qa/targets": 20702763319,
+    "ultra_qa/input_tokens": 20702803093,
+    "cortex/blocks": 12690000,
+    "cortex/targets": 482534935,
+    "cortex/input_tokens": 20171789678,
+    "ultra_style/blocks": 12690000,
+    "ultra_style/targets": 20070737053,
+    "ultra_style/input_tokens": 20071789036,
+    "openmath/blocks": 8460000,
+    "openmath/targets": 13072208273,
+    "openmath/input_tokens": 13076563771,
+    "rewrite6/blocks": 4230000,
+    "rewrite6/targets": 6934478168,
+    "rewrite6/input_tokens": 6942080796
   },
   "sources": {
     "climbmix": {
-      "page": 4609,
-      "row": 3963,
+      "page": 4634,
+      "row": 2400,
       "start": {
-        "file": 223,
+        "file": 224,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 9,
+          "row_group": 26,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 20480,
-      "documents": 18882427
+      "offset": 2048,
+      "documents": 18983264
     },
     "rewrite6": {
-      "page": 22,
-      "row": 2895,
+      "page": 36,
+      "row": 2703,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 90112
+          "row": 147456
         }
       },
       "epoch": 52,
       "offset": 0,
-      "documents": 10774015
+      "documents": 10831167
     },
     "openmath": {
-      "page": 2881,
-      "row": 724,
+      "page": 2896,
+      "row": 2245,
       "start": {
         "file": 27,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 11,
-          "row": 30
+          "row_group": 72,
+          "row": 470
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 11801300
+      "documents": 11864261
     },
     "ultra_qa": {
-      "page": 1245,
-      "row": 2808,
+      "page": 1252,
+      "row": 1383,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 35423
+          "row": 64095
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 5102328
+      "documents": 5129575
     },
     "ultra_style": {
-      "page": 1901,
-      "row": 453,
+      "page": 1911,
+      "row": 1420,
       "start": {
         "file": 11,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 2,
-          "row": 81501
+          "row_group": 3,
+          "row": 37668
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 7786949
+      "documents": 7828876
     },
     "cortex": {
-      "page": 244798,
-      "row": 127,
+      "page": 246106,
+      "row": 57,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 31334271
+      "documents": 31501625
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 935000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "32c5ec46c4d26de24f43e5e5ac475183bd0d5a9168b597e1a4995f06cda64d9b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c55cec8e778846c453320337a4022807364361a6bb535fd24e51ca5d5f695ab5", "training_config.json": "aafacc05aaf9ea874dcfa3d02457799efaf5525cf7b211a4208541f6e989efee", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "9525e5f4fb5139b10175901a17eb2c1ac2df509c25c3edcfbc3108aaabd5b46e", "coverage.json": "ea1e9229096e365e0bf25587fc7c00524d8f83b2754ab83b1c551cc02c49d653", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000935000"}
+{"step": 940000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "d09a6f6d15b58b397b5cb80bf0442935d3486b863565dc4a3b414d4d29102f73", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "379eeb0d627e11d6aeb818be90d2beb7ad27d2bcf192919683768be57619bc1b", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "e7f0f1e8787fc3b9f72fcf8f200650769ecf8ab43afd791855595c601ac93a27", "coverage.json": "3d6b4fdad9262d21f1491946476e23ba4ec13ac73d7d5a5a0af7a9f9856d8a06", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000940000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32c5ec46c4d26de24f43e5e5ac475183bd0d5a9168b597e1a4995f06cda64d9b
+oid sha256:d09a6f6d15b58b397b5cb80bf0442935d3486b863565dc4a3b414d4d29102f73
 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": 77571.26574998416,
-      "required_pages_per_second_upper_estimate": 0.005366560154718245
+      "required_token_positions_per_second": 77337.79198135664,
+      "required_pages_per_second_upper_estimate": 0.005350407897670814
     },
     {
       "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": 9696.40821874802,
-      "required_pages_per_second_upper_estimate": 0.004059683637160962
+      "required_token_positions_per_second": 9667.22399766958,
+      "required_pages_per_second_upper_estimate": 0.004047464813231256
     },
     {
       "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": 19392.81643749604,
-      "required_pages_per_second_upper_estimate": 0.0040390122409692635
+      "required_token_positions_per_second": 19334.44799533916,
+      "required_pages_per_second_upper_estimate": 0.004026855633747318
     },
     {
       "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": 29089.22465624406,
-      "required_pages_per_second_upper_estimate": 0.0017438546014675901
+      "required_token_positions_per_second": 29001.67199300874,
+      "required_pages_per_second_upper_estimate": 0.0017386059529918088
     },
     {
       "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": 29089.22465624406,
-      "required_pages_per_second_upper_estimate": 0.0022615592670458835
+      "required_token_positions_per_second": 29001.67199300874,
+      "required_pages_per_second_upper_estimate": 0.0022547524325828034
     },
     {
       "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": 29089.22465624406,
-      "required_pages_per_second_upper_estimate": 0.37312534031431177
+      "required_token_positions_per_second": 29001.67199300874,
+      "required_pages_per_second_upper_estimate": 0.37200230875705464
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,168 +155,30 @@
   "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": 64,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.05626329523441598,
-    "producer_p95_seconds": 0.1819149599996308,
-    "gpu_update_seconds": 0.9504550336670903,
-    "producer_batches_per_second": 17.773576820084738,
-    "gpu_batches_per_second": 1.052127627902346,
-    "producer_headroom": 16.892985554918255,
+    "producer_batch_seconds": 0.2906923827109864,
+    "producer_p95_seconds": 0.20526774199970532,
+    "gpu_update_seconds": 0.9533243464950897,
+    "producer_batches_per_second": 3.440062621091193,
+    "gpu_batches_per_second": 1.0489609372471331,
+    "producer_headroom": 3.2794954501539464,
     "sustainable": true,
-    "buffer_seconds": 60.82912215469378,
+    "buffer_seconds": 60.05943382919065,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 88,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 324038.0284101658,
-    "upload_bytes_per_second": 4252787.429657175,
+    "generation_bytes_per_second": 328751.58465276496,
+    "upload_bytes_per_second": 3596231.5696296524,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 69.53101767758112,
+    "estimated_full_shards_per_hour": 70.54243712126933,
     "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": 1.8764951709999877,
-      "bytes": 52123,
-      "pages": 1,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2044020090006597,
-      "bytes": 7888698,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2931400809993647,
-      "bytes": 6805061,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.581821284000398,
-      "bytes": 6273764,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9842661719994794,
-      "bytes": 2388376,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 4.260054792999654,
-      "bytes": 10427634,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4442174710002291,
-      "bytes": 7300139,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0513844460001565,
-      "bytes": 1119612,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 2.561698224999418,
-      "bytes": 16238211,
-      "pages": 36,
-      "shards": 4
-    },
-    {
-      "seconds": 2.406933221000145,
-      "bytes": 7260996,
-      "pages": 29,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1487480899995717,
-      "bytes": 1204399,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2970412000004217,
-      "bytes": 7490143,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 4.696426351999435,
-      "bytes": 5652066,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1291608219999034,
-      "bytes": 6292028,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.6897458180001195,
-      "bytes": 13896820,
-      "pages": 38,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1221129010000368,
-      "bytes": 948347,
-      "pages": 27,
-      "shards": 1
-    },
-    {
-      "seconds": 1.7077558099999806,
-      "bytes": 9679861,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4209248310007752,
-      "bytes": 7838674,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0323733719997108,
-      "bytes": 1179407,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2691897260001497,
-      "bytes": 7296014,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.574737728999935,
-      "bytes": 10770405,
-      "pages": 43,
-      "shards": 3
-    },
-    {
-      "seconds": 1.9674206870004127,
-      "bytes": 5752054,
-      "pages": 28,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3296988840002086,
-      "bytes": 7650823,
-      "pages": 38,
-      "shards": 3
-    },
     {
       "seconds": 1.2172214809997968,
       "bytes": 6191100,
@@ -460,6 +322,246 @@
       "bytes": 6201672,
       "pages": 37,
       "shards": 2
+    },
+    {
+      "seconds": 1.2297464679995755,
+      "bytes": 644185,
+      "pages": 19,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3486207620007917,
+      "bytes": 11139901,
+      "pages": 32,
+      "shards": 4
+    },
+    {
+      "seconds": 0.9246388850006042,
+      "bytes": 1100937,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.532587591000265,
+      "bytes": 13240452,
+      "pages": 31,
+      "shards": 4
+    },
+    {
+      "seconds": 1.8508207049999328,
+      "bytes": 10873151,
+      "pages": 40,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9900608629995986,
+      "bytes": 1198965,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.5980903330000729,
+      "bytes": 7181818,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9470139499990182,
+      "bytes": 1239114,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 2.040032099999735,
+      "bytes": 7666715,
+      "pages": 33,
+      "shards": 3
+    },
+    {
+      "seconds": 34.010384213999714,
+      "bytes": 15924428,
+      "pages": 39,
+      "shards": 4
+    },
+    {
+      "seconds": 1.0855836930004443,
+      "bytes": 1537634,
+      "pages": 45,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2260879469995416,
+      "bytes": 7238361,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2277038499996706,
+      "bytes": 10100946,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9283436340010667,
+      "bytes": 2568999,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.7634025150000525,
+      "bytes": 6258608,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2659589719987707,
+      "bytes": 11586393,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.253272633000961,
+      "bytes": 6708450,
+      "pages": 42,
+      "shards": 3
+    },
+    {
+      "seconds": 1.5571684259994072,
+      "bytes": 5848121,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1272584719990846,
+      "bytes": 2509117,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3743308060002164,
+      "bytes": 6082381,
+      "pages": 31,
+      "shards": 2
+    },
+    {
+      "seconds": 1.292268752000382,
+      "bytes": 11000537,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4258081890002359,
+      "bytes": 8234938,
+      "pages": 31,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4250404260001233,
+      "bytes": 6072497,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.7697906350003905,
+      "bytes": 6241903,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.7708508949999668,
+      "bytes": 6504956,
+      "pages": 35,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1802944610008126,
+      "bytes": 6153792,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.060665737999443,
+      "bytes": 2571654,
+      "pages": 33,
+      "shards": 2
+    },
+    {
+      "seconds": 1.9573821910016704,
+      "bytes": 16491610,
+      "pages": 35,
+      "shards": 4
+    },
+    {
+      "seconds": 1.2731761789982556,
+      "bytes": 6636284,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2044296969997959,
+      "bytes": 2640995,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.348795689000326,
+      "bytes": 5804474,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1589008739993005,
+      "bytes": 2402568,
+      "pages": 31,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2708631089990376,
+      "bytes": 9724430,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.795558247000372,
+      "bytes": 12019830,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9979082219997508,
+      "bytes": 2716683,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.3286046470002475,
+      "bytes": 11093643,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.2714691889996175,
+      "bytes": 7462349,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0150521480009047,
+      "bytes": 842077,
+      "pages": 26,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1759212130000378,
+      "bytes": 6164782,
+      "pages": 33,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4110855310009356,
+      "bytes": 8184458,
+      "pages": 31,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c55cec8e778846c453320337a4022807364361a6bb535fd24e51ca5d5f695ab5
-size 721765887
+oid sha256:379eeb0d627e11d6aeb818be90d2beb7ad27d2bcf192919683768be57619bc1b
+size 721778751

training_config.json

@@ -53,8 +53,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 64,
-  "prefetch_low_water": 32,
+  "prefetch_batches": 63,
+  "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,