Qwen90M step 545000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 540000,
+  "step": 545000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d",
+    "model.safetensors": "a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0",
-    "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
+    "trainer.pt": "abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7",
+    "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb",
     "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
-    "shard_scheduler.json": "de12fe13ce1f77cd03863bd480ed4c92361aad4e582dff7ec3d7a242555cb0c8",
-    "coverage.json": "c6a7d6cc92b6319e9b013aeaaaf6b0e9a72597bf84b77a74dadf487e898d80d9",
+    "shard_scheduler.json": "9362a296b2832b4b5007ca3ec1a25ba07f94fdb1146f86932cdf2269d63a6f4c",
+    "coverage.json": "682a00d4779d7723fea633cd98c833ad848bbb3df21fa445fa4095cd4ab9203f",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 540000,
+  "step": 545000,
   "stats": {
-    "climbmix/blocks": 19440000,
-    "climbmix/targets": 32495587833,
-    "climbmix/input_tokens": 32497646947,
-    "ultra_qa/blocks": 7290000,
-    "ultra_qa/targets": 11892698357,
-    "ultra_qa/input_tokens": 11892721312,
-    "cortex/blocks": 7290000,
-    "cortex/targets": 277235924,
-    "cortex/input_tokens": 11588137812,
-    "ultra_style/blocks": 7290000,
-    "ultra_style/targets": 11529724861,
-    "ultra_style/input_tokens": 11530329511,
-    "openmath/blocks": 4860000,
-    "openmath/targets": 7509510059,
-    "openmath/input_tokens": 7512011367,
-    "rewrite6/blocks": 2430000,
-    "rewrite6/targets": 3983637088,
-    "rewrite6/input_tokens": 3988004561
+    "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
   },
   "sources": {
     "climbmix": {
-      "page": 2662,
-      "row": 3055,
+      "page": 2687,
+      "row": 1151,
       "start": {
-        "file": 128,
+        "file": 130,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 68,
+          "row_group": 3,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 10906607
+      "offset": 2048,
+      "documents": 11007103
     },
     "rewrite6": {
-      "page": 14,
-      "row": 2627,
+      "page": 28,
+      "row": 3287,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 57344
+          "row": 114688
         }
       },
       "epoch": 30,
-      "offset": 2048,
-      "documents": 6222091
+      "offset": 0,
+      "documents": 6280095
     },
     "openmath": {
-      "page": 1663,
-      "row": 2767,
+      "page": 1679,
+      "row": 155,
       "start": {
         "file": 15,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 261,
-          "row": 898
+          "row_group": 327,
+          "row": 434
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 6814415
+      "documents": 6877339
     },
     "ultra_qa": {
-      "page": 719,
-      "row": 1872,
+      "page": 726,
+      "row": 497,
       "start": {
         "file": 5,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 5,
-          "row": 24371
+          "row": 53043
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 2946896
+      "offset": 4096,
+      "documents": 2974193
     },
     "ultra_style": {
-      "page": 1098,
-      "row": 616,
+      "page": 1108,
+      "row": 980,
       "start": {
         "file": 6,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 4,
-          "row": 47086
+          "row_group": 5,
+          "row": 2462
         }
       },
       "epoch": 0,
-      "offset": 4096,
-      "documents": 4498024
+      "offset": 2048,
+      "documents": 4539348
     },
     "cortex": {
-      "page": 141384,
-      "row": 37,
+      "page": 142692,
+      "row": 98,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 18097189
+      "documents": 18264674
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 540000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "de12fe13ce1f77cd03863bd480ed4c92361aad4e582dff7ec3d7a242555cb0c8", "coverage.json": "c6a7d6cc92b6319e9b013aeaaaf6b0e9a72597bf84b77a74dadf487e898d80d9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000540000"}
+{"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"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d
+oid sha256:a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b
 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": 76448.66075247723,
-      "required_pages_per_second_upper_estimate": 0.005288895736188249
+      "required_token_positions_per_second": 77724.66568028349,
+      "required_pages_per_second_upper_estimate": 0.00537717271783323
     },
     {
       "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": 9556.082594059653,
-      "required_pages_per_second_upper_estimate": 0.004000932228436205
+      "required_token_positions_per_second": 9715.583210035436,
+      "required_pages_per_second_upper_estimate": 0.004067711805593652
     },
     {
       "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": 19112.165188119307,
-      "required_pages_per_second_upper_estimate": 0.00398055998699525
+      "required_token_positions_per_second": 19431.166420070873,
+      "required_pages_per_second_upper_estimate": 0.00404699953098255
     },
     {
       "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": 28668.24778217896,
-      "required_pages_per_second_upper_estimate": 0.0017186176806618552
+      "required_token_positions_per_second": 29146.74963010631,
+      "required_pages_per_second_upper_estimate": 0.0017473031358150829
     },
     {
       "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": 28668.24778217896,
-      "required_pages_per_second_upper_estimate": 0.0022288301667688993
+      "required_token_positions_per_second": 29146.74963010631,
+      "required_pages_per_second_upper_estimate": 0.002266031580738054
     },
     {
       "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": 28668.24778217896,
-      "required_pages_per_second_upper_estimate": 0.3677255009835554
+      "required_token_positions_per_second": 29146.74963010631,
+      "required_pages_per_second_upper_estimate": 0.37386320891351205
     }
   ],
   "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": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.3263542902969334,
-    "producer_p95_seconds": 0.2078722389996983,
-    "gpu_update_seconds": 0.9644119239539579,
-    "producer_batches_per_second": 3.064154600480815,
-    "gpu_batches_per_second": 1.0369013231401534,
-    "producer_headroom": 2.9551072335420745,
+    "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,
     "sustainable": true,
-    "buffer_seconds": 60.757951209099346,
+    "buffer_seconds": 60.709067819084495,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 393197.65477415506,
-    "upload_bytes_per_second": 5782840.946028825,
+    "generation_bytes_per_second": 297407.2945755357,
+    "upload_bytes_per_second": 5461795.829792473,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 84.37106354158868,
+    "estimated_full_shards_per_hour": 63.81668212842515,
     "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": 1.4530193070004316,
-      "bytes": 7690650,
-      "pages": 38,
+      "seconds": 1.1747847249989718,
+      "bytes": 10716254,
+      "pages": 34,
       "shards": 3
     },
     {
-      "seconds": 1.0416680760008603,
-      "bytes": 5628205,
-      "pages": 24,
+      "seconds": 1.0711626960001013,
+      "bytes": 2631436,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.0915457699993567,
-      "bytes": 13214903,
-      "pages": 39,
-      "shards": 4
-    },
-    {
-      "seconds": 1.4333472569996957,
-      "bytes": 6210705,
-      "pages": 36,
+      "seconds": 1.137612402000741,
+      "bytes": 5722821,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.232096203000765,
-      "bytes": 1201273,
-      "pages": 35,
-      "shards": 1
-    },
-    {
-      "seconds": 1.2199994860002334,
-      "bytes": 7355179,
-      "pages": 26,
+      "seconds": 1.4303736129986646,
+      "bytes": 11919616,
+      "pages": 42,
       "shards": 3
     },
     {
-      "seconds": 1.1187948159986263,
-      "bytes": 10435269,
-      "pages": 45,
+      "seconds": 1.0499967649993778,
+      "bytes": 7310677,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 0.8605066399995849,
-      "bytes": 757336,
-      "pages": 24,
-      "shards": 1
+      "seconds": 1.0670534790006059,
+      "bytes": 5802765,
+      "pages": 28,
+      "shards": 2
     },
     {
-      "seconds": 1.4440002779992938,
-      "bytes": 13477323,
-      "pages": 45,
-      "shards": 4
+      "seconds": 0.9111343770000531,
+      "bytes": 2649643,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 1.0930595740010176,
-      "bytes": 6005049,
-      "pages": 30,
+      "seconds": 1.2196053769985156,
+      "bytes": 6057143,
+      "pages": 37,
       "shards": 2
     },
     {
-      "seconds": 1.1409170160004578,
-      "bytes": 7751245,
-      "pages": 38,
+      "seconds": 1.2347321990000637,
+      "bytes": 9643472,
+      "pages": 35,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1648727040010272,
+      "bytes": 8469052,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.385997141000189,
-      "bytes": 5854774,
+      "seconds": 1.1223673130007228,
+      "bytes": 6068911,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1270621679996111,
+      "bytes": 5842224,
       "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 0.8962630369987892,
-      "bytes": 1124581,
-      "pages": 35,
-      "shards": 1
+      "seconds": 1.3927226420000807,
+      "bytes": 7163586,
+      "pages": 36,
+      "shards": 4
     },
     {
-      "seconds": 1.1323460120001982,
-      "bytes": 7094094,
-      "pages": 25,
-      "shards": 3
+      "seconds": 1.230830958000297,
+      "bytes": 6020552,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.3368760009998368,
-      "bytes": 10726853,
+      "seconds": 1.2729082169989852,
+      "bytes": 7700586,
       "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 0.9546346269999049,
-      "bytes": 5999756,
-      "pages": 29,
+      "seconds": 0.9891643420014589,
+      "bytes": 6883017,
+      "pages": 33,
       "shards": 2
     },
     {
-      "seconds": 1.1835868019988993,
-      "bytes": 7840439,
-      "pages": 38,
+      "seconds": 1.2779713299987634,
+      "bytes": 9455856,
+      "pages": 30,
       "shards": 3
     },
     {
-      "seconds": 1.00542208799925,
-      "bytes": 5360697,
-      "pages": 33,
+      "seconds": 1.159703276000073,
+      "bytes": 2570411,
+      "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2967056479992607,
-      "bytes": 7612762,
-      "pages": 38,
-      "shards": 3
+      "seconds": 1.2064628600001015,
+      "bytes": 6189523,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9550784620005288,
-      "bytes": 883731,
+      "seconds": 1.4404870660000597,
+      "bytes": 11907122,
+      "pages": 40,
+      "shards": 4
+    },
+    {
+      "seconds": 0.842472636999446,
+      "bytes": 857193,
       "pages": 26,
       "shards": 1
     },
     {
-      "seconds": 1.1562814480002999,
-      "bytes": 11922123,
-      "pages": 40,
+      "seconds": 1.0425618379995285,
+      "bytes": 11991809,
+      "pages": 35,
       "shards": 3
     },
     {
-      "seconds": 1.2267546230013977,
-      "bytes": 11778233,
-      "pages": 33,
-      "shards": 4
+      "seconds": 1.543527111000003,
+      "bytes": 7770797,
+      "pages": 39,
+      "shards": 3
     },
     {
-      "seconds": 0.9305929850015673,
-      "bytes": 1085853,
-      "pages": 35,
+      "seconds": 1.0592705330000172,
+      "bytes": 5143840,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.4772485220000817,
-      "bytes": 7565247,
-      "pages": 36,
-      "shards": 3
+      "seconds": 0.9495372719993611,
+      "bytes": 5544847,
+      "pages": 18,
+      "shards": 2
     },
     {
-      "seconds": 1.035932449000029,
-      "bytes": 5952411,
-      "pages": 30,
+      "seconds": 1.068828384999506,
+      "bytes": 2455135,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 0.7514245589991333,
-      "bytes": 257606,
-      "pages": 8,
-      "shards": 1
+      "seconds": 1.2464688429990929,
+      "bytes": 6118587,
+      "pages": 36,
+      "shards": 2
     },
     {
-      "seconds": 0.9524619530002383,
-      "bytes": 4638620,
-      "pages": 26,
-      "shards": 2
+      "seconds": 0.8108600380001008,
+      "bytes": 790179,
+      "pages": 24,
+      "shards": 1
     },
     {
-      "seconds": 1.3739831450002384,
-      "bytes": 13357834,
-      "pages": 40,
-      "shards": 4
+      "seconds": 2.05223027700049,
+      "bytes": 18642145,
+      "pages": 45,
+      "shards": 5
     },
     {
-      "seconds": 1.1055001610002364,
-      "bytes": 6014129,
+      "seconds": 1.3427877739995893,
+      "bytes": 6044460,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 0.9146980179993989,
-      "bytes": 2516554,
-      "pages": 34,
+      "seconds": 1.2501696180006547,
+      "bytes": 2425262,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.1637235170001077,
-      "bytes": 10485207,
-      "pages": 34,
-      "shards": 3
-    },
-    {
-      "seconds": 1.0128666689997772,
-      "bytes": 6455974,
+      "seconds": 1.180065475000447,
+      "bytes": 6094205,
       "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.0892136950005806,
-      "bytes": 8273419,
-      "pages": 32,
+      "seconds": 1.2719860980032536,
+      "bytes": 10158142,
+      "pages": 41,
       "shards": 3
     },
     {
-      "seconds": 1.3893709189997026,
-      "bytes": 6184789,
-      "pages": 36,
+      "seconds": 0.8902074819998234,
+      "bytes": 2207572,
+      "pages": 24,
       "shards": 2
     },
     {
-      "seconds": 1.226723518999279,
-      "bytes": 5211595,
-      "pages": 30,
+      "seconds": 1.290630097999383,
+      "bytes": 12177821,
+      "pages": 42,
+      "shards": 3
+    },
+    {
+      "seconds": 3.9100393570006418,
+      "bytes": 2234803,
+      "pages": 29,
       "shards": 2
     },
     {
-      "seconds": 1.4460434569991776,
-      "bytes": 7613200,
-      "pages": 39,
+      "seconds": 0.9967152519966476,
+      "bytes": 11168227,
+      "pages": 38,
       "shards": 3
     },
     {
-      "seconds": 1.303638638000848,
-      "bytes": 6027470,
-      "pages": 37,
+      "seconds": 1.0345216240020818,
+      "bytes": 6241760,
+      "pages": 36,
       "shards": 2
     },
     {
-      "seconds": 1.9445103569996718,
-      "bytes": 2479005,
-      "pages": 31,
+      "seconds": 1.262870265996753,
+      "bytes": 2380360,
+      "pages": 28,
       "shards": 2
     },
     {
-      "seconds": 1.5074067269997613,
-      "bytes": 15537281,
-      "pages": 35,
-      "shards": 4
-    },
-    {
-      "seconds": 0.9092938880003203,
-      "bytes": 1113077,
-      "pages": 34,
-      "shards": 1
+      "seconds": 0.9234551270019438,
+      "bytes": 5872826,
+      "pages": 30,
+      "shards": 2
     },
     {
-      "seconds": 1.0412252780006384,
-      "bytes": 7665353,
-      "pages": 39,
+      "seconds": 1.268712075001531,
+      "bytes": 10759229,
+      "pages": 40,
       "shards": 3
     },
     {
-      "seconds": 1.1747847249989718,
-      "bytes": 10716254,
-      "pages": 34,
+      "seconds": 1.2720383849991777,
+      "bytes": 7578314,
+      "pages": 39,
       "shards": 3
     },
     {
-      "seconds": 1.0711626960001013,
-      "bytes": 2631436,
-      "pages": 37,
-      "shards": 2
-    },
-    {
-      "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,
+      "seconds": 1.0708622599995579,
+      "bytes": 1033241,
+      "pages": 32,
+      "shards": 1
+    },
+    {
+      "seconds": 1.2405112279993773,
+      "bytes": 12607883,
+      "pages": 39,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1709360159984499,
+      "bytes": 6311701,
       "pages": 35,
       "shards": 2
     },
     {
-      "seconds": 1.2064628600001015,
-      "bytes": 6189523,
-      "pages": 36,
+      "seconds": 1.1361314889982168,
+      "bytes": 6780795,
+      "pages": 32,
       "shards": 2
     },
     {
-      "seconds": 1.4404870660000597,
-      "bytes": 11907122,
-      "pages": 40,
-      "shards": 4
+      "seconds": 1.0505489870010933,
+      "bytes": 7688939,
+      "pages": 38,
+      "shards": 3
     },
     {
-      "seconds": 0.842472636999446,
-      "bytes": 857193,
-      "pages": 26,
-      "shards": 1
+      "seconds": 1.4415530749975005,
+      "bytes": 9558241,
+      "pages": 36,
+      "shards": 3
     },
     {
-      "seconds": 1.0425618379995285,
-      "bytes": 11991809,
-      "pages": 35,
-      "shards": 3
+      "seconds": 2.3017471879975346,
+      "bytes": 2579509,
+      "pages": 37,
+      "shards": 2
     },
     {
-      "seconds": 1.543527111000003,
-      "bytes": 7770797,
-      "pages": 39,
-      "shards": 3
+      "seconds": 1.0516134489989781,
+      "bytes": 6165714,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 0.908222386999114,
+      "bytes": 5648395,
+      "pages": 35,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0
-size 721757375
+oid sha256:abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7
+size 721774207

training_config.json

@@ -46,8 +46,8 @@
   "cpu_threads": 4,
   "page_rows": 4096,
   "cortex_page_rows": 128,
-  "prefetch_batches": 63,
-  "prefetch_low_water": 31,
+  "prefetch_batches": 64,
+  "prefetch_low_water": 32,
   "memory_pages": 12,
   "cache_gb": 20,
   "force_rebenchmark": false,