Update TensorBoard training metrics

benchmark_cache/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": 76747.2747425519,
-      "required_pages_per_second_upper_estimate": 0.005309554544901533
+      "required_token_positions_per_second": 76953.78316171891,
+      "required_pages_per_second_upper_estimate": 0.005323841276505055
     },
     {
       "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": 9593.409342818988,
-      "required_pages_per_second_upper_estimate": 0.004016560158670589
+      "required_token_positions_per_second": 9619.222895214863,
+      "required_pages_per_second_upper_estimate": 0.00402736775401047
     },
     {
       "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": 19186.818685637976,
-      "required_pages_per_second_upper_estimate": 0.0039961083417832674
+      "required_token_positions_per_second": 19238.445790429727,
+      "required_pages_per_second_upper_estimate": 0.004006860906213079
     },
     {
       "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": 28780.228028456964,
-      "required_pages_per_second_upper_estimate": 0.0017253307254422856
+      "required_token_positions_per_second": 28857.66868564459,
+      "required_pages_per_second_upper_estimate": 0.0017299731745956469
     },
     {
       "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": 28780.228028456964,
-      "required_pages_per_second_upper_estimate": 0.0022375361383679647
+      "required_token_positions_per_second": 28857.66868564459,
+      "required_pages_per_second_upper_estimate": 0.00224355680883885
     },
     {
       "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": 28780.228028456964,
-      "required_pages_per_second_upper_estimate": 0.36916186334778883
+      "required_token_positions_per_second": 28857.66868564459,
+      "required_pages_per_second_upper_estimate": 0.37015518894889227
     }
   ],
   "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": 704566,
-    "producer_batch_seconds": 0.3711980840780029,
-    "producer_p95_seconds": 0.3522699639943312,
-    "gpu_update_seconds": 0.9606595184952165,
-    "producer_batches_per_second": 2.6939794220216444,
-    "gpu_batches_per_second": 1.0409515345940743,
-    "producer_headroom": 2.587996974395334,
+    "producer_batch_seconds": 0.14911970157817223,
+    "producer_p95_seconds": 0.3527131480004755,
+    "gpu_update_seconds": 0.9580815519499555,
+    "producer_batches_per_second": 6.706022003911906,
+    "gpu_batches_per_second": 1.0437524842898072,
+    "producer_headroom": 6.424915968918469,
     "sustainable": true,
-    "buffer_seconds": 60.521549665198634,
+    "buffer_seconds": 60.3591377728472,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 434873.4045947675,
+    "generation_bytes_per_second": 433595.8650456332,
     "upload_bytes_per_second": 5060189.13756225,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 93.31370929128903,
+    "estimated_full_shards_per_hour": 93.03957904364346,
     "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": [

metrics/void-byte-v1/1790868774355361490-de221e1d/train.jsonl

@@ -379,3 +379,4 @@
 {"step": 478100, "loss": 0.5561420321464539, "lr": 0.0003, "grad_norm": 0.1717190543169397, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 69978879133, "tokens_per_second": 152332.04869220534, "data_wait_s": 1.1302996426820755e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 478200, "loss": 0.6358839273452759, "lr": 0.0003, "grad_norm": 0.15193957228271637, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 69993405592, "tokens_per_second": 151187.67555844056, "data_wait_s": 1.780599995981902e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 478300, "loss": 0.5612347722053528, "lr": 0.0003, "grad_norm": 0.1444592162545219, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 70007921025, "tokens_per_second": 149904.0171321939, "data_wait_s": 1.1409996659494936e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 478400, "loss": 0.5991029143333435, "lr": 0.0003, "grad_norm": 0.1899080482951865, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 70022483508, "tokens_per_second": 152213.58236681233, "data_wait_s": 1.0787996870931238e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790868774.f92f6aaf-743a-4c24-a54a-850bd75954cb-8js8b.612.0.1790868774355361490-de221e1d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:90d6c198215966b77061f5d2f5e099faab938565100242612d6670197af93fe4
-size 812604
+oid sha256:7c6eeefa9c75450f0d4bfde4baf229119dba798bca3d327945a4036cc150a85a
+size 814722