Update TensorBoard training metrics

benchmark_cache/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": 74315.68651620558,
-      "required_pages_per_second_upper_estimate": 0.006441495355839216
+      "required_token_positions_per_second": 74744.57936571156,
+      "required_pages_per_second_upper_estimate": 0.006478670701015405
     },
     {
       "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": 9289.460814525697,
-      "required_pages_per_second_upper_estimate": 0.003705704808730532
+      "required_token_positions_per_second": 9343.072420713945,
+      "required_pages_per_second_upper_estimate": 0.003727091280003967
     },
     {
       "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": 18578.921629051394,
-      "required_pages_per_second_upper_estimate": 0.004297906541028049
+      "required_token_positions_per_second": 18686.14484142789,
+      "required_pages_per_second_upper_estimate": 0.004322710744147265
     },
     {
       "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": 27868.38244357709,
-      "required_pages_per_second_upper_estimate": 0.0016970382767111847
+      "required_token_positions_per_second": 28029.217262141832,
+      "required_pages_per_second_upper_estimate": 0.0017068322733266993
     },
     {
       "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": 27868.38244357709,
-      "required_pages_per_second_upper_estimate": 0.0021453852251266435
+      "required_token_positions_per_second": 28029.217262141832,
+      "required_pages_per_second_upper_estimate": 0.0021577667346790305
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "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": 27868.38244357709,
-      "required_pages_per_second_upper_estimate": 0.34603204047303837
+      "required_token_positions_per_second": 28029.217262141832,
+      "required_pages_per_second_upper_estimate": 0.34802907064010125
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.2398038003830152,
-    "producer_p95_seconds": 0.30634939200172084,
-    "gpu_update_seconds": 0.9920920260074914,
-    "producer_batches_per_second": 4.170075696893868,
-    "gpu_batches_per_second": 1.0079710085205835,
-    "producer_headroom": 4.137098846736039,
+    "producer_batch_seconds": 0.0707526302890642,
+    "producer_p95_seconds": 0.23438371700103744,
+    "gpu_update_seconds": 0.9863992897633738,
+    "producer_batches_per_second": 14.133750164685592,
+    "gpu_batches_per_second": 1.013788240094829,
+    "producer_headroom": 13.941521124138836,
     "sustainable": true,
-    "buffer_seconds": 60.517613586456974,
+    "buffer_seconds": 60.1703566755658,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 411473.90439870243,
+    "generation_bytes_per_second": 411380.66500180477,
     "upload_bytes_per_second": 5223183.501047112,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 88.29272126169973,
+    "estimated_full_shards_per_hour": 88.27271425762756,
     "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-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -214,3 +214,4 @@
 {"step": 61400, "loss": 0.6789006367325783, "lr": 0.0003, "grad_norm": 0.15025199092677508, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 8986351874, "tokens_per_second": 151812.90684847214, "data_wait_s": 9.141000191448256e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 61500, "loss": 0.6323178894817829, "lr": 0.0003, "grad_norm": 0.1411816376063164, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9001065635, "tokens_per_second": 149132.58962462685, "data_wait_s": 1.1607000487856567e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 61600, "loss": 0.5627754870802164, "lr": 0.0003, "grad_norm": 0.13681460468315645, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9015734789, "tokens_per_second": 150014.01704744526, "data_wait_s": 1.049899947247468e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 61700, "loss": 0.627738730981946, "lr": 0.0003, "grad_norm": 0.1597955363305401, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9030170596, "tokens_per_second": 146411.2573257933, "data_wait_s": 8.319002517964691e-06, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f0654f0558e2040a136cfd49e3f06a4b2374e8de3523586f29b340f4dd48c7f2
-size 460785
+oid sha256:5dbef9fe24c407082fac75d56c07b5d8317a5ebdcad7de1b11546831de3ae2c5
+size 462903