Qwen90M step 160000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 155000,
+  "step": 160000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c",
+    "model.safetensors": "204af5881a577fef1ed88738397873be9f9b5acc4b3f00f0f4a334c0cc5e8b87",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725",
-    "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
-    "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4",
-    "shard_scheduler.json": "81a7ca71b967fe09918b8d8fc0d0910018e4a0b46e1c1831d67cb3e3d678dfdb",
-    "coverage.json": "66656093e66bc4071d1aa3216cfbf69575a2e82a31d04761ddf384991f4d7c54",
+    "trainer.pt": "58b1a9a87c98368688872d521c56ed9c3b9c63110540b20f692c5f45af15e7f4",
+    "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
+    "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
+    "shard_scheduler.json": "b3d4d07ae7a6200c6ddee288c163b45048e3d78bd520197a93226391eb377b08",
+    "coverage.json": "a820d638cc498caf3588ee86be26b0f7f1fac193033602fe4336d6fc6fd2fc93",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 155000,
+  "step": 160000,
   "stats": {
-    "climbmix/blocks": 5580000,
-    "climbmix/targets": 9325510201,
-    "climbmix/input_tokens": 9326101685,
-    "ultra_qa/blocks": 2092500,
-    "ultra_qa/targets": 3413363253,
-    "ultra_qa/input_tokens": 3413369829,
-    "cortex/blocks": 2092500,
-    "cortex/targets": 79585819,
-    "cortex/input_tokens": 3325972346,
-    "ultra_style/blocks": 2092500,
-    "ultra_style/targets": 3309559754,
-    "ultra_style/input_tokens": 3309733622,
-    "openmath/blocks": 1395000,
-    "openmath/targets": 2155399566,
-    "openmath/input_tokens": 2156118041,
-    "rewrite6/blocks": 697500,
-    "rewrite6/targets": 1143387147,
-    "rewrite6/input_tokens": 1144640945
+    "climbmix/blocks": 5760000,
+    "climbmix/targets": 9627080225,
+    "climbmix/input_tokens": 9627690635,
+    "ultra_qa/blocks": 2160000,
+    "ultra_qa/targets": 3523611388,
+    "ultra_qa/input_tokens": 3523618170,
+    "cortex/blocks": 2160000,
+    "cortex/targets": 82159978,
+    "cortex/input_tokens": 3433330864,
+    "ultra_style/blocks": 2160000,
+    "ultra_style/targets": 3416345189,
+    "ultra_style/input_tokens": 3416524370,
+    "openmath/blocks": 1440000,
+    "openmath/targets": 2224889648,
+    "openmath/input_tokens": 2225631152,
+    "rewrite6/blocks": 720000,
+    "rewrite6/targets": 1180373319,
+    "rewrite6/input_tokens": 1181667487
   },
   "sources": {
     "climbmix": {
-      "page": 764,
-      "row": 3782,
+      "page": 789,
+      "row": 1412,
       "start": {
-        "file": 36,
+        "file": 38,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 74,
+          "row_group": 10,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 3133126
+      "offset": 0,
+      "documents": 3233156
     },
     "rewrite6": {
-      "page": 34,
-      "row": 3725,
+      "page": 48,
+      "row": 3853,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 139264
+          "row": 196608
         }
       },
       "epoch": 8,
-      "offset": 0,
-      "documents": 1786221
+      "offset": 2048,
+      "documents": 1843693
     },
     "openmath": {
-      "page": 477,
-      "row": 2923,
+      "page": 493,
+      "row": 341,
       "start": {
         "file": 4,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 207,
-          "row": 192
+          "row_group": 272,
+          "row": 728
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 1956715
+      "offset": 2048,
+      "documents": 2019669
     },
     "ultra_qa": {
-      "page": 206,
-      "row": 2172,
+      "page": 213,
+      "row": 777,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 1242
+          "row": 29914
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 845948
+      "documents": 873225
     },
     "ultra_style": {
-      "page": 315,
-      "row": 1608,
+      "page": 325,
+      "row": 1744,
       "start": {
         "file": 1,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 7,
-          "row": 9777
+          "row": 50737
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 1291848
+      "offset": 0,
+      "documents": 1332944
     },
     "cortex": {
-      "page": 40579,
-      "row": 14,
+      "page": 41888,
+      "row": 1,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 5194126
+      "documents": 5361665
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 155000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4", "shard_scheduler.json": "81a7ca71b967fe09918b8d8fc0d0910018e4a0b46e1c1831d67cb3e3d678dfdb", "coverage.json": "66656093e66bc4071d1aa3216cfbf69575a2e82a31d04761ddf384991f4d7c54", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000155000"}
+{"step": 160000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "204af5881a577fef1ed88738397873be9f9b5acc4b3f00f0f4a334c0cc5e8b87", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "58b1a9a87c98368688872d521c56ed9c3b9c63110540b20f692c5f45af15e7f4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "b3d4d07ae7a6200c6ddee288c163b45048e3d78bd520197a93226391eb377b08", "coverage.json": "a820d638cc498caf3588ee86be26b0f7f1fac193033602fe4336d6fc6fd2fc93", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000160000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:118aac7ab3812af64a497800a8642841ab877140d3f257367ab42e73fd8e186c
+oid sha256:204af5881a577fef1ed88738397873be9f9b5acc4b3f00f0f4a334c0cc5e8b87
 size 360622744

performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 86,
     "mode": "compiled",
-    "microbatch_seconds": 0.8971913670000049,
+    "microbatch_seconds": 0.8980338759999995,
     "peak_gb": 85.39913082122803,
     "fits_budget": true,
-    "estimated_update_seconds": 0.9531432340000379,
-    "input_tokens_per_second": 196310.40431087767
+    "estimated_update_seconds": 0.9548798630000306,
+    "input_tokens_per_second": 196126.2316567667
   },
   "measurements": [
     {

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 74435.92703712951,
-      "required_pages_per_second_upper_estimate": 0.006451917499446089
+      "required_token_positions_per_second": 75895.57656905257,
+      "required_pages_per_second_upper_estimate": 0.006578436221425263
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9304.49087964119,
-      "required_pages_per_second_upper_estimate": 0.0037117005264246006
+      "required_token_positions_per_second": 9486.947071131572,
+      "required_pages_per_second_upper_estimate": 0.0037844850291732773
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18608.98175928238,
-      "required_pages_per_second_upper_estimate": 0.0043048604231167645
+      "required_token_positions_per_second": 18973.894142263143,
+      "required_pages_per_second_upper_estimate": 0.0043892764806807705
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 27913.47263892357,
-      "required_pages_per_second_upper_estimate": 0.0016997840330378133
+      "required_token_positions_per_second": 28460.841213394717,
+      "required_pages_per_second_upper_estimate": 0.0017331159074021418
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 27913.47263892357,
-      "required_pages_per_second_upper_estimate": 0.002148856393182065
+      "required_token_positions_per_second": 28460.841213394717,
+      "required_pages_per_second_upper_estimate": 0.002190994341258046
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.042780299000014566,
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
-      "estimated_pages_for_buffer": 21,
-      "required_token_positions_per_second": 27913.47263892357,
-      "required_pages_per_second_upper_estimate": 0.3465919097920654
+      "estimated_pages_for_buffer": 22,
+      "required_token_positions_per_second": 28460.841213394717,
+      "required_pages_per_second_upper_estimate": 0.35338839556222257
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,413 +155,125 @@
   "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": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 61,
+    "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.1047992101799764,
-    "producer_p95_seconds": 0.4328458870004397,
-    "gpu_update_seconds": 0.9904894442064731,
-    "producer_batches_per_second": 9.54205664606303,
-    "gpu_batches_per_second": 1.0096018749610667,
-    "producer_headroom": 9.451306383945653,
+    "producer_batch_seconds": 0.5166383134453412,
+    "producer_p95_seconds": 0.1841241980000632,
+    "gpu_update_seconds": 0.9714400144640783,
+    "producer_batches_per_second": 1.935590090737235,
+    "gpu_batches_per_second": 1.0293996387946585,
+    "producer_headroom": 1.8803096657423064,
     "sustainable": true,
-    "buffer_seconds": 60.41985609659486,
+    "buffer_seconds": 60.229280896772856,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 76,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 341840.10677463334,
-    "upload_bytes_per_second": 5861021.62228704,
+    "generation_bytes_per_second": 394137.7785946498,
+    "upload_bytes_per_second": 4508842.591861955,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 73.35092928342104,
+    "estimated_full_shards_per_hour": 84.57279222850438,
     "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": 1.2166951500003051,
-      "bytes": 7600114,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.056514903997595,
-      "bytes": 5902748,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0406464490006329,
-      "bytes": 6219293,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1162217770033749,
-      "bytes": 6495200,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1027901979978196,
-      "bytes": 11991791,
-      "pages": 37,
-      "shards": 3
-    },
-    {
-      "seconds": 1.5872819810028886,
-      "bytes": 6214149,
-      "pages": 24,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0351565999953891,
-      "bytes": 6216009,
-      "pages": 41,
-      "shards": 2
-    },
-    {
-      "seconds": 1.4164216759963892,
-      "bytes": 2450413,
-      "pages": 33,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0167613629964762,
-      "bytes": 6214181,
-      "pages": 34,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3737143940015812,
-      "bytes": 10463664,
-      "pages": 36,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8019685460021719,
-      "bytes": 2186825,
-      "pages": 23,
-      "shards": 2
-    },
-    {
-      "seconds": 1.2952469239971833,
-      "bytes": 12034549,
-      "pages": 44,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8924442770003225,
-      "bytes": 6155396,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.1965257189949625,
-      "bytes": 6621394,
-      "pages": 31,
-      "shards": 4
-    },
-    {
-      "seconds": 1.1455287000062526,
-      "bytes": 6484150,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 0.8978742559993407,
-      "bytes": 2461928,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9092525360028958,
-      "bytes": 6015320,
-      "pages": 26,
-      "shards": 2
-    },
-    {
-      "seconds": 1.3533267960010562,
-      "bytes": 11703451,
-      "pages": 42,
-      "shards": 3
-    },
-    {
-      "seconds": 1.2227107450016774,
-      "bytes": 7523214,
-      "pages": 33,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0604035779979313,
-      "bytes": 6119324,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0826924209977733,
-      "bytes": 902486,
-      "pages": 27,
+      "seconds": 1.0642098359999181,
+      "bytes": 45649,
+      "pages": 1,
       "shards": 1
     },
     {
-      "seconds": 1.1300200109981233,
-      "bytes": 7391590,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8869208329997491,
-      "bytes": 4199574,
-      "pages": 23,
-      "shards": 2
-    },
-    {
-      "seconds": 1.295291658003407,
-      "bytes": 11877354,
-      "pages": 32,
+      "seconds": 1.2323853809998582,
+      "bytes": 7278265,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.0650672470001155,
-      "bytes": 2271023,
-      "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0174418120004702,
-      "bytes": 5974488,
-      "pages": 31,
-      "shards": 2
-    },
-    {
-      "seconds": 1.083592351002153,
-      "bytes": 5253625,
-      "pages": 35,
-      "shards": 2
-    },
-    {
-      "seconds": 1.5616185160033638,
-      "bytes": 7297224,
+      "seconds": 0.9598245239999414,
+      "bytes": 1062466,
       "pages": 33,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1017810449993704,
-      "bytes": 6135514,
-      "pages": 36,
-      "shards": 2
-    },
-    {
-      "seconds": 0.7622847379971063,
-      "bytes": 678366,
-      "pages": 20,
       "shards": 1
     },
     {
-      "seconds": 1.3286253929982195,
-      "bytes": 17203756,
-      "pages": 37,
-      "shards": 5
-    },
-    {
-      "seconds": 0.8924150489983731,
-      "bytes": 1133711,
+      "seconds": 1.1868087180000657,
+      "bytes": 11845509,
       "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.4622295160006615,
-      "bytes": 7475361,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0509287640015827,
-      "bytes": 1010925,
-      "pages": 30,
-      "shards": 1
-    },
-    {
-      "seconds": 0.9923871680002776,
-      "bytes": 11139165,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 1.103713971999241,
-      "bytes": 7399212,
-      "pages": 30,
-      "shards": 3
+      "shards": 4
     },
     {
-      "seconds": 1.030579199003114,
-      "bytes": 6815769,
+      "seconds": 0.9347497089997887,
+      "bytes": 6852819,
       "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 1.1106542109992006,
-      "bytes": 7819068,
-      "pages": 40,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8756177200048114,
-      "bytes": 4958523,
-      "pages": 27,
+      "seconds": 1.0938459449998845,
+      "bytes": 6142145,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 0.9241842089977581,
-      "bytes": 5682376,
+      "seconds": 1.2391148739998243,
+      "bytes": 7141661,
       "pages": 29,
-      "shards": 2
-    },
-    {
-      "seconds": 0.9258055570026045,
-      "bytes": 7457751,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 1.6118608709948603,
-      "bytes": 885264,
-      "pages": 28,
-      "shards": 1
-    },
-    {
-      "seconds": 1.013624004997837,
-      "bytes": 9878612,
-      "pages": 14,
-      "shards": 3
-    },
-    {
-      "seconds": 1.218867665003927,
-      "bytes": 8704871,
-      "pages": 46,
       "shards": 3
     },
     {
-      "seconds": 1.0288945650027017,
-      "bytes": 5962293,
-      "pages": 29,
+      "seconds": 3.5350073540002995,
+      "bytes": 5274454,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 0.9867427620047238,
-      "bytes": 6108091,
-      "pages": 38,
+      "seconds": 1.0228857370002515,
+      "bytes": 6079721,
+      "pages": 34,
       "shards": 2
     },
     {
-      "seconds": 0.9743170239962637,
-      "bytes": 2340987,
-      "pages": 25,
-      "shards": 3
-    },
-    {
-      "seconds": 1.4316346049963613,
-      "bytes": 10012608,
+      "seconds": 1.587895904999641,
+      "bytes": 7946846,
       "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 0.8118417049990967,
-      "bytes": 973375,
-      "pages": 27,
-      "shards": 1
+      "seconds": 1.0177612440002122,
+      "bytes": 6784132,
+      "pages": 31,
+      "shards": 2
     },
     {
-      "seconds": 1.5195651320027537,
-      "bytes": 13317753,
-      "pages": 32,
+      "seconds": 1.3712029609996534,
+      "bytes": 12398330,
+      "pages": 38,
       "shards": 4
     },
     {
-      "seconds": 1.0066122520001954,
-      "bytes": 1096551,
+      "seconds": 0.7755766290001702,
+      "bytes": 1144894,
       "pages": 33,
       "shards": 1
     },
     {
-      "seconds": 1.1023049610012094,
-      "bytes": 10542150,
-      "pages": 30,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9890626239939593,
-      "bytes": 7730044,
-      "pages": 39,
-      "shards": 3
-    },
-    {
-      "seconds": 0.8456406199984485,
-      "bytes": 955055,
-      "pages": 28,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2024492920027114,
-      "bytes": 7626352,
-      "pages": 38,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9993227560044033,
-      "bytes": 10754654,
-      "pages": 25,
-      "shards": 3
-    },
-    {
-      "seconds": 0.9575367360012024,
-      "bytes": 6150393,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "seconds": 1.0901530560004176,
-      "bytes": 2315430,
-      "pages": 30,
-      "shards": 2
-    },
-    {
-      "seconds": 0.974858529996709,
-      "bytes": 6294959,
-      "pages": 37,
+      "seconds": 1.4222684020005545,
+      "bytes": 6265294,
+      "pages": 27,
       "shards": 2
     },
     {
-      "seconds": 1.211790827997902,
-      "bytes": 10339350,
-      "pages": 32,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1235615409968887,
-      "bytes": 2445045,
-      "pages": 36,
+      "seconds": 0.9120372190000126,
+      "bytes": 2656908,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.1649940130009782,
-      "bytes": 11872287,
-      "pages": 31,
-      "shards": 3
-    },
-    {
-      "seconds": 1.1487145990031422,
-      "bytes": 2742520,
-      "pages": 38,
+      "seconds": 1.6420339440001044,
+      "bytes": 5755818,
+      "pages": 29,
       "shards": 2
-    },
-    {
-      "seconds": 1.1007023089987342,
-      "bytes": 9561100,
-      "pages": 32,
-      "shards": 3
     }
   ],
   "cache_key": {
@@ -606,7 +318,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    2.5993740069999944,
-    0.014666526000041813
+    2.7487001069999906,
+    0.015045025999995687
   ]
 }

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:26893e627deb9ceeba478e819739d4b42e9e7f80f0f5fa2f1efa5408ac2b8725
-size 721762815
+oid sha256:58b1a9a87c98368688872d521c56ed9c3b9c63110540b20f692c5f45af15e7f4
+size 721773183

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 61,
-  "prefetch_low_water": 30,
+  "prefetch_batches": 62,
+  "prefetch_low_water": 31,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,