Qwen90M step 810000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 805000,
+  "step": 810000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "f81353344f2caa3305d0563c6e2992bc9eca917fdb0f44f55a65bbf690de87f9",
+    "model.safetensors": "b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "1232789bccab1559e521ddbb828d7a33d03698d8a2d36aeb0341e6a2e27d4c92",
+    "trainer.pt": "a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb",
     "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
-    "shard_scheduler.json": "671cb6a206e18282c9946eb8a1b247f024ce1af7927d526a978429fd995d40c4",
-    "coverage.json": "413a5d11bd5c4ca98af7d6cdbc7f9d3951a2feb0417fce131129dc6f92afcc10",
+    "shard_scheduler.json": "d3fc1112c30fbcd245d56533cfa5eb69c877b6ac83ebcc2954147694a2f34a08",
+    "coverage.json": "284e9a6216e7c4741fa1f7e01c706d28eb52bfba351b9120d3853f5e4505d59b",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 805000,
+  "step": 810000,
   "stats": {
-    "climbmix/blocks": 28980000,
-    "climbmix/targets": 48444394688,
-    "climbmix/input_tokens": 48447464623,
-    "ultra_qa/blocks": 10867500,
-    "ultra_qa/targets": 17729141745,
-    "ultra_qa/input_tokens": 17729175912,
-    "cortex/blocks": 10867500,
-    "cortex/targets": 413246363,
-    "cortex/input_tokens": 17274632186,
-    "ultra_style/blocks": 10867500,
-    "ultra_style/targets": 17189552467,
-    "ultra_style/input_tokens": 17190453888,
-    "openmath/blocks": 7245000,
-    "openmath/targets": 11194750151,
-    "openmath/input_tokens": 11198479888,
-    "rewrite6/blocks": 3622500,
-    "rewrite6/targets": 5938628484,
-    "rewrite6/input_tokens": 5945139165
+    "climbmix/blocks": 29160000,
+    "climbmix/targets": 48745844785,
+    "climbmix/input_tokens": 48748933709,
+    "ultra_qa/blocks": 10935000,
+    "ultra_qa/targets": 17839350326,
+    "ultra_qa/input_tokens": 17839384717,
+    "cortex/blocks": 10935000,
+    "cortex/targets": 415808421,
+    "cortex/input_tokens": 17381926974,
+    "ultra_style/blocks": 10935000,
+    "ultra_style/targets": 17296108093,
+    "ultra_style/input_tokens": 17297015211,
+    "openmath/blocks": 7290000,
+    "openmath/targets": 11264198535,
+    "openmath/input_tokens": 11267951300,
+    "rewrite6/blocks": 3645000,
+    "rewrite6/targets": 5975471100,
+    "rewrite6/input_tokens": 5982022644
   },
   "sources": {
     "climbmix": {
-      "page": 3969,
-      "row": 558,
+      "page": 3993,
+      "row": 2480,
       "start": {
-        "file": 192,
+        "file": 193,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 9,
+          "row_group": 23,
           "row": 1024
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 16257582
+      "documents": 16357808
     },
     "rewrite6": {
-      "page": 7,
-      "row": 3564,
+      "page": 22,
+      "row": 256,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 28672
+          "row": 90112
         }
       },
       "epoch": 45,
-      "offset": 0,
-      "documents": 9275416
+      "offset": 2048,
+      "documents": 9333548
     },
     "openmath": {
-      "page": 2480,
-      "row": 1695,
+      "page": 2495,
+      "row": 3276,
       "start": {
         "file": 23,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 115,
-          "row": 130
+          "row_group": 176,
+          "row": 570
         }
       },
       "epoch": 0,
-      "offset": 0,
-      "documents": 10159775
+      "offset": 2048,
+      "documents": 10222796
     },
     "ultra_qa": {
-      "page": 1072,
-      "row": 2161,
+      "page": 1079,
+      "row": 780,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 3,
-          "row": 40360
+          "row_group": 4,
+          "row": 4352
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 4393073
+      "offset": 4096,
+      "documents": 4420364
     },
     "ultra_style": {
-      "page": 1636,
-      "row": 3207,
+      "page": 1647,
+      "row": 173,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 6,
-          "row": 27157
+          "row": 72213
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 6704263
+      "documents": 6746285
     },
     "cortex": {
-      "page": 210760,
-      "row": 21,
+      "page": 212069,
+      "row": 58,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 26977301
+      "documents": 27144890
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 805000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f81353344f2caa3305d0563c6e2992bc9eca917fdb0f44f55a65bbf690de87f9", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1232789bccab1559e521ddbb828d7a33d03698d8a2d36aeb0341e6a2e27d4c92", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "671cb6a206e18282c9946eb8a1b247f024ce1af7927d526a978429fd995d40c4", "coverage.json": "413a5d11bd5c4ca98af7d6cdbc7f9d3951a2feb0417fce131129dc6f92afcc10", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000805000"}
+{"step": 810000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "d3fc1112c30fbcd245d56533cfa5eb69c877b6ac83ebcc2954147694a2f34a08", "coverage.json": "284e9a6216e7c4741fa1f7e01c706d28eb52bfba351b9120d3853f5e4505d59b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000810000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f81353344f2caa3305d0563c6e2992bc9eca917fdb0f44f55a65bbf690de87f9
+oid sha256:b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4
 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": 76909.91644078335,
-      "required_pages_per_second_upper_estimate": 0.005320806474965917
+      "required_token_positions_per_second": 76742.70439241717,
+      "required_pages_per_second_upper_estimate": 0.0053092383574745395
     },
     {
       "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": 9613.739555097918,
-      "required_pages_per_second_upper_estimate": 0.004025071994008668
+      "required_token_positions_per_second": 9592.838049052147,
+      "required_pages_per_second_upper_estimate": 0.0040163209699003825
     },
     {
       "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": 19227.479110195836,
-      "required_pages_per_second_upper_estimate": 0.004004576835931166
+      "required_token_positions_per_second": 19185.676098104293,
+      "required_pages_per_second_upper_estimate": 0.0039958703709320606
     },
     {
       "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": 28841.218665293753,
-      "required_pages_per_second_upper_estimate": 0.001728987020993341
+      "required_token_positions_per_second": 28778.514147156442,
+      "required_pages_per_second_upper_estimate": 0.0017252279808752647
     },
     {
       "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": 28841.218665293753,
-      "required_pages_per_second_upper_estimate": 0.0022422778921125652
+      "required_token_positions_per_second": 28778.514147156442,
+      "required_pages_per_second_upper_estimate": 0.002237402891635416
     },
     {
       "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": 28841.218665293753,
-      "required_pages_per_second_upper_estimate": 0.36994418575048743
+      "required_token_positions_per_second": 28778.514147156442,
+      "required_pages_per_second_upper_estimate": 0.3691398795186881
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.0739909046796825,
-    "producer_p95_seconds": 0.15531819999978325,
-    "gpu_update_seconds": 0.9586280080900458,
-    "producer_batches_per_second": 13.515174660036216,
-    "gpu_batches_per_second": 1.0431575038083678,
-    "producer_headroom": 12.95602496333958,
+    "producer_batch_seconds": 0.1648581179296542,
+    "producer_p95_seconds": 0.19809514199914702,
+    "gpu_update_seconds": 0.9607167298014186,
+    "producer_batches_per_second": 6.065822008393333,
+    "gpu_batches_per_second": 1.0408895452530542,
+    "producer_headroom": 5.8275366834611155,
     "sustainable": true,
-    "buffer_seconds": 60.393564509672885,
+    "buffer_seconds": 60.525153977489374,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 81,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 383905.7653317095,
-    "upload_bytes_per_second": 3532908.8334232424,
+    "generation_bytes_per_second": 375296.30858700874,
+    "upload_bytes_per_second": 4130574.7776954207,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 82.37724037135567,
+    "estimated_full_shards_per_hour": 80.52985137183853,
     "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": [
@@ -226,6 +226,246 @@
       "bytes": 2500711,
       "pages": 36,
       "shards": 2
+    },
+    {
+      "seconds": 1.54583506400013,
+      "bytes": 5956633,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.6070335669992346,
+      "bytes": 5667674,
+      "pages": 32,
+      "shards": 2
+    },
+    {
+      "seconds": 1.649759572000221,
+      "bytes": 7616382,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.8874110320002728,
+      "bytes": 11916705,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.604149986000266,
+      "bytes": 2375336,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4722619869999107,
+      "bytes": 10022494,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.460268009000174,
+      "bytes": 1130080,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.6416546800001015,
+      "bytes": 7689453,
+      "pages": 38,
+      "shards": 4
+    },
+    {
+      "seconds": 1.5363667999999961,
+      "bytes": 6120972,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.693875957000273,
+      "bytes": 10707952,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.540582020000329,
+      "bytes": 7589375,
+      "pages": 37,
+      "shards": 3
+    },
+    {
+      "seconds": 1.573228470999311,
+      "bytes": 6089737,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 1.5443801999999778,
+      "bytes": 7496481,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.4706297409993567,
+      "bytes": 5704544,
+      "pages": 26,
+      "shards": 2
+    },
+    {
+      "seconds": 1.637408425999638,
+      "bytes": 1096275,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.9842258310000034,
+      "bytes": 13303551,
+      "pages": 43,
+      "shards": 4
+    },
+    {
+      "seconds": 1.4946416839993617,
+      "bytes": 5892242,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 1.5706974580007227,
+      "bytes": 6564398,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.5560058639985073,
+      "bytes": 6086581,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.8728917230000661,
+      "bytes": 6075781,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.5521995129984134,
+      "bytes": 2545582,
+      "pages": 34,
+      "shards": 2
+    },
+    {
+      "seconds": 1.8264212399990356,
+      "bytes": 16670948,
+      "pages": 38,
+      "shards": 4
+    },
+    {
+      "seconds": 1.714815281000483,
+      "bytes": 7600635,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3434111860005942,
+      "bytes": 1056341,
+      "pages": 35,
+      "shards": 1
+    },
+    {
+      "seconds": 1.5977993900014553,
+      "bytes": 5977865,
+      "pages": 28,
+      "shards": 2
+    },
+    {
+      "seconds": 1.8666853409995383,
+      "bytes": 6845301,
+      "pages": 39,
+      "shards": 3
+    },
+    {
+      "seconds": 1.52379110199945,
+      "bytes": 6381300,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4397832380000182,
+      "bytes": 6693273,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 1.380750825999712,
+      "bytes": 7212989,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.7552309799993964,
+      "bytes": 9939009,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.423224260999632,
+      "bytes": 2660337,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.6043403389994637,
+      "bytes": 6329322,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.7717889699997613,
+      "bytes": 5364175,
+      "pages": 27,
+      "shards": 2
+    },
+    {
+      "seconds": 1.5835829299994657,
+      "bytes": 13733184,
+      "pages": 40,
+      "shards": 4
+    },
+    {
+      "seconds": 1.5670746849991701,
+      "bytes": 6380319,
+      "pages": 29,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1554255430000921,
+      "bytes": 1163617,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 2.120087691999288,
+      "bytes": 11751236,
+      "pages": 36,
+      "shards": 4
+    },
+    {
+      "seconds": 1.3820675099996151,
+      "bytes": 1056686,
+      "pages": 32,
+      "shards": 1
+    },
+    {
+      "seconds": 1.7487962389986933,
+      "bytes": 7557748,
+      "pages": 38,
+      "shards": 4
+    },
+    {
+      "seconds": 1.7842456250000396,
+      "bytes": 12131753,
+      "pages": 31,
+      "shards": 3
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1232789bccab1559e521ddbb828d7a33d03698d8a2d36aeb0341e6a2e27d4c92
-size 721769471
+oid sha256:a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb
+size 721768959